1. Lots of news reports emerged recently suggesting China under-reporting or faking COVID19 cases. Unfortunately, most of these reports did not discuss much scientific evidence. This long thread is my attempt to discuss existing scientific evidence on this issue. https://twitter.com/BBCWorld/status/1247480464442626048

2. Based on existing evidence, my overall assessment is that case detection in China is at least comparable or better than most more developed countries. There's no solid evidence suggesting severe under-reporting. See this thread for details.

3. First of all, as Dr. @mlipsitch pointed out, true infection number is higher than reported number. This is true for any country, and is caused by things like limited testing capacity, testing criteria, spectrum of disease severity etc. https://twitter.com/mlipsitch/status/1239173318638534657

4. Surveillance typically biases towards detecting clinically severe cases. After all, hospitals cannot detect someone who has too mild a symptom to go to the hospital. When testing is limited, it is often prioritized for severe cases and mild case are more likely to be missed.

5. Based on this bit of data from @WHO on the spectrum of #COVID19 severity, we can roughly expect that around 18%-60% cases of all infection can be detected. This is a ballpark estimate because of the uncertainty about the spectrum of severity. https://twitter.com/HelenBranswell/status/1225802856478822400

6. What % of infections were detected in China? To answer this, we need to know the total number of infections. Using cases detected internationally and air traffic data, the Imperial College team estimated 4000 cases with symptom onset by 1/18 in Wuhan.

https://www.imperial.ac.uk/media/imperial-college/medicine/mrc-gida/2020-01-22-COVID19-Report-2.pdf

7. But not all travel related cases are detected. @NeneRiehus et al. estimated that ~38% are detected globally relative to Singapore's surveillance. That leads to at least 4000/0.38=10526 with symptom onset by 1/18 in Wuhan. https://www.thelancet.com/pdfs/journals/laninf/PIIS1473-3099(20)30229-2.pdf