Back in early Feb, we suggested the COVID outbreak was about to peak in Wuhan (bloomberg.com/news/articles/…). Looking at the data now, surely this was obvious? A brief thread on real-time analysis and managing hindsight... 1/
When we suggested the outbreak was peaking, there was a lot of scepticism, including from other modelling groups (e.g. nature.com/articles/d4158…). Reported cases seemed to be slowing down, but maybe testing was at capacity and growth in infections was going undetected? 2/
On 13th Feb, there was a huge spike in cases as China changed their case definition. Try and put hindsight to one side and read this article as if in real-time. What would your conclusion about the China outbreak be? theguardian.com/world/2020/feb… 3/ 
At the time, we were revising a paper on early COVID dynamics. Our working conclusion was that there had been a decline in transmission in late Jan, coinciding with control measures. But had we got it wrong? Was the outbreak actually still accelerating? 4/
We reanalysed the model and available data, and concluded that on the weight of evidence, there had been a decline (since published here: thelancet.com/journals/lanin…). Later in Feb, the WHO joint mission report confirmed evidence of this decline: who.int/docs/default-s… 5/
To my knowledge, we were 1st modelling group to publicly call the peak in Wuhan. But I’m not telling this story because it was easy & obvious & everyone else should have seen it. Although our conclusion turned out to be on right lines, it was anything but obvious at the time. 6/
So here’s a suggestion. Draw some lines on this picture about what you think might happen over the coming months. Then, in future, look back on what you thought now, what has changed, and what it tells you about your real-time perspective. 7/7
