Thread by @globalhlthtwit

Covid UK: Different models make different assumptions. 'Clearly, there is a substantial amount of uncertainty about these long-term forecasts as indicated by the wide credible intervals. Part of this uncertainty is due to the fluctuations in transmission risk and mobility.' (1)

Here is the LSHTM model for a large surge in hospitalisations in September 2021 (2)

Here is the Imperial model for hospital admissions. Looks a bit worse, with an earlier even higher peak under the worst case scenario (3)

And the Warwick model which has its peak earlier in August, and a worst case of almost double the hospital admissions peak in January. For all models see gov.uk/government/pub… (4)

UCL's Friston dynamic causal model is different. It responds "to real-time deaths, infections, vaccinations, vaccine effectiveness data, hospital admissions + Google mobility trends, informs transmission trends + the 'R' rate, now estimated as 1.12." (5) telegraph.co.uk/news/2021/04/0…

Critics point out that Friston's model was overly optimistic last year, which he agrees with, but also partly because of the emergence of the new UK variant. But he says that the new model is informed by better current data.(6)

I'm not a modeller. I can see arguments for pessimism about another surge, + also for vaccines bringing about herd immunity quicker than we hoped. Either way, we need to focus on vaxx AND the public health response to lock down infections + contacts as quickly as possible. (7)

If you want to watch two animated simulations of natural spread of the virus and the impact of rapid testing and isolation on spread, watch Nobel laureate Paul Romer's two models from a year ago. paulromer.net/covid-sim-part… and paulromer.net/covid-sim-part… (8)

Local public health teams (1 worker per 1500 pop) w finance support, are the best way to identify cases and contacts quickly, + to monitor and ensure compliance with isolation. Asian states look on with shock at the failure of the UK, USA and most European states to do this. (9)