Making predictions about Covid-19

A University of Sussex researcher is contributing to the development of Covid-19 forecasting reports which are being used by leading public health organisations, including the World Health Organisation and the US Centers for Disease Control and Prevention, the main public health institute of the United States, to guide public policy and better understand how the emerging disease is affecting countries around the world.

Covid-19 virus

“Mathematical models are useful for putting pieces of information together in a coherent framework,” said Dr Pierre Nouvellet, from the School of Life Sciences, University of Sussex talking about the Covid-19 weekly reports he has helped an international team of epidemiologists produce since early April. “For instance, it’s very clear that interventions around social distancing are quite efficient,” Nouvellet said. “We see clear patterns around the time that the social distancing measures are implemented. If you wait for the delay, around 15 days, then you can see a clear decrease in the number of deaths. This pattern is shared among many countries.”

Nouvellet is part of a group which produces leading weekly reports which forecast the reported number of Covid-19 related deaths in the week ahead and whether country Covid-19 transmissions are declining or increasing in countries around the world. The predicted models are being used by public health organisations, such as the US Centers for Disease Control and Prevention, the leading national public health institute in America, and the World Health Organisation, and by many health ministries around the world, who need to make difficult decisions around how to plan for enough hospital beds, protective equipment or staff to treat potential future patients. The models, which have received extensive media coverage, are also one of the 21 models being used by the well-known University of Massachusetts’ Reich Lab, a leading flu forecasting centre in America, which is now forecasting the spread of Covid-19 in that country.

“During an outbreak of an epidemic disease, it’s quite hard to keep up with the situation and to have a good assessment of what is happening,” said Nouvellet. “So the forecasts and methods that we are using are tools for situation awareness. We know from feedback, they are helpful for public health officials to understand what the situation is, and also the impact of the interventions they are putting in place, to see if they are working. I wouldn’t say that they drive policy makers, but they do inform them.”

The group is made up of individuals and teams from Sussex, the Jameel Institute (J-IDEA) at Imperial, the WHO Collaborating Centre for Infectious Disease Modelling and the MRC Centre for Global Infectious Disease Analysis at Imperial, which Nouvellet was part of for 5 years. The group meets daily online to present and scrutinise results, as part of an informal peer review.

Nouvellet has developed statistical methods which integrate various sources of information to understand the transmission of infectious and emerging diseases for many years. Prior to joining Sussex in 2017, he worked at the MRC Centre for Global Infectious Disease Analysis at Imperial on outbreak responses to the MERS-coronavirus in Saudi Arabia and to Ebola in both west Africa and the Democratic Republic of Congo. He now heads up the Infectious Diseases Modelling Group within the School of Life Sciences at the University of Sussex.

Making predictions about the future

Making projections about the future, and especially about a future that has changed so dramatically in the past few months, is a very complex task. So how do they do it? The group only releases forecasts for one week into the future, because factors, such as whether policies around a lock down are lifted, would dramatically change the projections. Nouvellet and Dr Sangeeta Bhatia, Research Associate in Infectious Disease modelling at Imperial, who are leading the development of the weekly forecast reports, combine predictions from four different mathematical models to get estimates that they are confident in.

“As it’s in the future, we can’t observe Covid-19. We can only form theories about how it will be. That’s why we …use four models, because none of the models can capture the entire, full picture,” said Bhatia. Three of the models they use were developed by Nouvellet and Bhatia, and the fourth was developed by another team from the MRC Centre for Global Infectious Disease Analysis led by Dr Samir Bhatt. “We have combined the predictions from all of them because they will give us a range of possible scenarios that will happen in the next week.”

Despite the rigorous mathematical models, uncertainty remains. The forecasts use the number of deaths reported by Health Ministries from around the world to the European Centre for Disease Prevention and Control, an agency of the European Union. The forecasts use Covid-19 reported deaths because they are more likely to be reported than cases, which might be missed or under-reported, and which varies from country to country and over time depending on policies and testing capacity. The forecasts include data from any country that has had at least 100 reported deaths from Covid-19 so far, and at least ten observed deaths in each of the past two weeks. At the time of publication, China wasn’t included in the latest forecast because it no longer meets the criteria due to a reduction in its reported deaths.

“The full impact of Covid-19 is not represented”

But even this data has its weaknesses. Current UK data is based mostly on reported deaths in hospital, and so won’t include other deaths, such as in care homes. “We are clearly underestimating the number of deaths,” says Nouvellet, when we speak.

“We are really worried that the full impact of Covid-19 is not represented in these numbers,” said Bhatia. “There is a general concern, not only in the UK but also in low and middle-income countries, that deaths from Covid-19 are being under-reported. To get a good picture of how Covid-19 is impacting on countries it would be very helpful to get better data on deaths and cases.”

The reports also forecast Covid-19 transmissibility, the likelihood of it being passed on from one person to another, based on the reproduction number: the average number of people that one infected individual is likely to infect with Covid-19. Nouvellet says the forecasts on Covid-19 transmissibility should be accurate as they are based on consistent data, even if it only covers say 50% of recorded deaths. “The mean reproduction number estimated would be the same, even if we knew about every single death, but under reporting leads to increased uncertainty,” Nouvellet said. At the time of publication, the UK’s reproduction number is estimated at 0.84, meaning every person infected is infecting less than 1 person on average, and the reports forecast a decrease in number of reported deaths in the coming week. Upcoming forecasts will also include mobility data, location data that Apple and Google maps and tracks, which may more clearly explain how people’s movement correlates with infections and the reproduction rate.

“The Covid-19 pandemic is probably the most traumatic thing that has happened to us in our lifetime,” said Bhatia. “To not fully understand the scope of the pandemic is very unnerving. Our forecasts aim to provide some clarity – that’s the hope anyway.”

Contact us

Research development enquiries:
researchexternal@sussex.ac.uk

Research impact enquiries:
rqi@sussex.ac.uk

Research governance enquiries:
rgoffice@sussex.ac.uk

Doctoral study enquiries:
doctoralschool@sussex.ac.uk

Undergraduate research enquiries:
undergraduate-research@sussex.ac.uk

General press enquiries: 
press@sussex.ac.uk