Paperonium

Infectious disease · Nat Commun · 2020

0.31% of recovered COVID patients retested positive, range 0.42–0.57%, which misses 0.31%. Its count gives 0.25–0.38%.

The summarymisses 0.31%Its own count: 107 of 34,424contains 0.31%0.31%0.423 to 0.574%should be inside0.25 to 0.38%0.2%0.3%0.4%0.5%0.6%
The summarymisses 0.31%Its own count:107 of 34,424contains 0.31%0.31%0.423 to 0.574%should be inside0.25 to 0.38%0.2%0.4%0.6%

Percent of recovered patients retesting positive

The paper says

Wuhan tested nearly 10 million people for COVID after lockdown. Of 34,424 who had recovered from it, 107 tested positive again: 0.31%. The paper’s summary, the part most people read, gives this rate a range of 0.42% to 0.57%.

Its own data show

That range doesn’t contain 0.31%. The method the paper names, applied to its own count, 107 of 34,424, gives 0.25% to 0.38%; five other standard methods agree, and none reaches 0.42%. The same method exactly reproduces 27 of its other printed ranges, for test-positive rates by age, sex and district, and antibody results.

A range is the span of true rates the data can’t rule out, built around the measured rate, so it must contain it. The 0.31% itself fits the counts, and the paper’s conclusion that the rate is low stands. But a review pooling many studies’ ranges would place this study near 0.50%, the printed range’s middle, not 0.31%.

“The summary” means the paper’s abstract: the short summary printed at its top, which is the part most readers see.

Read the paper →Full technical write-up →

Post-lockdown SARS-CoV-2 nucleic acid screening in nearly ten million residents of Wuhan, China · Nat Commun (2020) · doi:10.1038/s41467-020-19802-w