Thread by @Ayjchan

Dec 2019 to mid-Jan 2020: Chinese CDC, Hubei CDC, Wuhan CDC specifically looked for potential Covid-19 cases with links to Huanan Market or living in the vicinity of the market.

Feb 2022: Western scientists say, "Wow so many of the early cases were centered around the market!"

China-WHO report annexes (p125) described the early search for cases:
"screening.. targeting people with pneumonia.. and exposure history with Huanan market.. surveillance at several hospitals (close to Huanan market), Huanan market and the neighbourhood"
who.int/publications/i…

Jan 2020, the 2019-nCoV Outbreak Joint Field Epidemiology Investigation Team reported that, in late Dec 2019, Wuhan CDC did "a retrospective search for pneumonia patients potentially linked to the market.. found additional patients linked to the market"
ncbi.nlm.nih.gov/labs/pmc/artic…

Feb 2022: Western scientists (Proximal Origin authors and friends) write that the "epidemiological link to the Huanan market is genuine and not due to ascertainment bias caused by special focus on the market as a possible site of cases"
zenodo.org/record/6299116…

Note that the targeted search for Covid-19 cases linked to the market or living near the market had begun upon the Wuhan CDC hearing that 4 cases had been detected at the market.

4 cases.
ncbi.nlm.nih.gov/labs/pmc/artic…

Models are only as good as the data they're trained on.

If the data is poor quality (e.g., data does not even exist, only has low n) and is biased, then the model and its predictions - regardless of p-value - are also poor quality and biased.

https://twitter.com/BiophysicsFL/status/1498335727305863172

Can someone explain to me how you can model an outbreak using just this picture from the China-WHO study without access to the data or methods used to plot this figure in the first place?

Is there a new scientific dark arts I haven't heard of?

That low resolution figure from the China-WHO report was transformed into this figure in the recent Worobey et al. preprint.

See informative thread by @danwalker9999 on problems with simplifying early Covid-19 case home addresses into contour plots.

https://twitter.com/danwalker9999/status/1499228514532270083?s=20&t=kuJTxTnlX4RITFxY2zquZw

If you want details, see page 26 Worobey preprint.

In sum: we don't have the data, we don't know how the China-WHO low res maps of early cases were generated, but we stacked 2 of these on top of each other, manually scaled one of them, and then reverse engineered data points.

The desperate data-less approach here is still undermined by the fact that most of the early cases were identified by targeting the market and the surrounding neighborhood.

The authors shouldn't have been surprised that cases clustered around the market.

https://twitter.com/Ayjchan/status/1499466077075644422

The Chinese investigators behaved as one would expect. After hearing of 4 cases from the market, they checked nearby hospitals & neighborhood for cases with potential links to the market.

So when you look at the identified early cases, it looks like the market is the epicenter.

We can say that the first 4 market cases were identified without a bias, but every case identified after that - with or without a link to the market - was likely affected by ascertainment bias among the Wuhan CDC investigators.

Remember they didn't have rapid diagnostics then.

On Jan 3, 2020, the Chinese authorities instructed early patient samples to be destroyed so one could not even retrospectively check if these were false positives or negatives.

h/t @Daoyu15
newsweek.com/china-acknowle…

@Daoyu15 Another scientist told me that the phrase I'm looking for is "downsampling until only the bias is left".

@Daoyu15 Had to share this by @gdemaneuf ... a few of the early Covid-19 cases appear to be living on water bodies or on the edge of water bodies.

On a serious note, it's likely the data points were jittered or lost resolution during image (re-re-re)processing.

https://twitter.com/gdemaneuf/status/1499259279161049090?s=20&t=DoQPoZeVJdGFSzfaHNU4gw

@Daoyu15 @gdemaneuf A good find by @TheSeeker268: doctors in Wuhan said that unless pneumonia cases were linked to the Huanan market, they did not meet Wuhan Municipal Health Commission standards and were often not reported or counted. archive.ph/iMQVD

One ICU director ignored the Huanan criteria & reported cases, but none were classified as novel coronavirus cases.

Many of his cases were not linked to the market. Some had been wandering outside for more than a week.
很多患者没有华南海鲜市场接触史
已经在外面游荡了一个多星期

There were quick consequences for this doctor who ignored the Huanan market link requirement. A leader at his hospital severely criticized him, saying he had reported too much.
他无视“标准”上报病例的事情很快有了回响，是来自一位院领导的“严厉批评”，“嫌我们报太多”。

Aside from the fact that early cases were reported/identified with a significant skew towards the Huanan market + nearby hospitals + its neighborhood... it's really tragic that this market-centric criteria prevented people from receiving healthcare.

A lot of people in the US are frustrated that they didn't get to be tested for Covid in early 2020.

Imagine being the people in Wuhan who were living in the outbreak area and yet couldn't get diagnosed/treated for Covid because they had no link to the downtown seafood market.