Imperial College of London COVID-19 forecasting models
The MRC Centre for Global Infectious Disease Analysis at Imperial College of London captured the world's attention with their report in March 2020 from their first COVID-19 model that projected that the British healthcare system and healthcare systems around the world would be overwhelmed with coronavirus patients unless countries immediately took aggressive measures to slow the spread of the virus.[1]
The world faces a severe and acute public health emergency due to the ongoing COVID-19 global pandemic. How individual countries respond in the coming weeks will be critical in influencing the trajectory of national epidemics.
Their first COVID-19 model is an existing model built in 2005 to study H5N1 avian flu that they adapted to model SARS-CoV-2.[2] They have since added a second model, a Bayesian statistical
model similar to the models at IHME, the University of Texas, and PolicyLab.
What does that first model say?
The key conclusion of the model that has had a big impact is the estimation that COVID-19 could kill 40 million people worldwide, without any intervention. And that the number of deaths could be cut in half through mitigation to slow the spread of the virus.
We estimate that in the absence of interventions, COVID-19 would have resulted in 7.0 billion infections and 40 million deaths globally this year. Mitigation strategies focussing on shielding the elderly (60% reduction in social contacts) and slowing but not interrupting transmission (40% reduction in social contacts for wider population) could reduce this burden by half, saving 20 million lives, but we predict that even in this scenario, health systems in all countries will be quickly overwhelmed.
The report concludes the need to begin suppression strategies immediately in multiple countries in order to save tens of millions of lives.
The new model from Imperial College
The early COVID-19 models didn't have a fine level of granularity. The first
model from Imperial College issued reports at the country level. It looked
at the entire UK, and the entire United States, as one area.
Imperial College has since responded to the call to provided finer-grained forecasting for making local decisions by creating a second model that can generate projections at the state level.[3] The new model is a Bayesian statistical model, unlike their first model that was an epidemological model.[4] The new model features a dashboard that shows a US overview, and a page for each US state. Each state includes projections for infections, deaths, and also an estimate for the current reprodution number.
The Imperial College COVID-19 Response Team has apparently learned from some of the criticism that Neil Ferguson endured for not immediately providing the source code for their original epidemiological model, and they provided the source code for their new model along with their first published report based on its projections.
References
1. Walker PGT, Whittaker C, Watson O, et al. Report 12: The Global Impact of COVID-19 and Strategies for Mitigation and Suppression. Imperial College London - MRC Centre for Global Infectious Disease Analysis; 26-April-2020 link; DOI: 10.25561/77735 (Accessed 2020-04-19 11:30)
2. Special report: The simulations driving the world’s response to COVID-19. Nature; 2-April-2020 link (Accessed 2020-04-20 19:30)
3. COVID-19 model introduction. Imperial College COVID-19 Response Team; link (Accessed 2020-05-24 21:30)
4. Unwin HJT, Mishra∗2 S, Bradley VC, et al. Report 23: State-level tracking of COVID-19 in the United States. Imperial College COVID-19 Response Team; 21-May-2020 link; DOI: 10.25561/79231
