Read on Twitter
Professor Karl Friston is one of the top neuroscience modellers in the world. In this TWO hour video and paper he explains his new COVID-19 model, now sent to SAGE. As a non-mathematician I understood about 50% of it, but learned some lessons. https://www.fil.ion.ucl.ac.uk/spm/covid-19/ (1)
He uses an SEIR model (susceptible, exposed, infected and recovered) with four states based on your location parameters, your infection parameters, clinical parameters and testing status.
He explains the components of his model: generative, prior, likelihood, free energy, hierarchical, dynamic, general linear and with a soupcon of Bayesian Model reduction to eliminate redundant parameters. Good for mathematicians, tricky for me. (3)
He has some unknown parameters (4)
And he does a number of cross-country comparisons based on the findings. A few points of note:
1. Germany keeps people alive longer mainly because of better acute care in hospital. Spain does badly on this count.
2. France is bad at social distancing, Canada is good.
1. Germany keeps people alive longer mainly because of better acute care in hospital. Spain does badly on this count.
2. France is bad at social distancing, Canada is good.
3. France is better at reducing contacts when they do manage social distancing.
4. UK does very badly at reducing the effective number of contacts in the home and at work so the effect of social distancing is markedly attenuated.
4. UK does very badly at reducing the effective number of contacts in the home and at work so the effect of social distancing is markedly attenuated.
5. In fact, Friston believes UK SD has almost no impact on illness and death rates, but works by sharing the virus more effectively in the home and rapidly building herd immunity!!
6. He sees the UK as a series of city epidemics, each of which in turn builds herd immunity.
6. He sees the UK as a series of city epidemics, each of which in turn builds herd immunity.
7. This has implications for the regional relaxation of SD and whether transport across regions should be limited.
8. Here is his narrative conclusion for London.
8. Here is his narrative conclusion for London.
9. I will be more than interested to know what SAGE and other modellers think of this new model. He said little about the importance of mass testing, but the need for effective contact tracing and emphasis on reducing contact at home seems important.
10. He seems to support the Oxford modellers in terms of herd immunity, that this will be inevitable and high as the epidemic spreads. Does he make too many assumptions about immunity here? If he's right, there will be no second wave in London.
