Day 82: Retroactive Deaths and Mutant Strains

SFGate reports on three deaths in the San Francisco area recently attributed to COVID-19 through autopsy. Two of them predeceased the first official American death in Kirkland, WA on February 28th, having died at home in Santa Clara County on February 6th and 17th, respectively. The third retroactive death occurred on March 6th, three days before the first known COVID-19 death in the county on March 9th. Neither the article nor the press release reveals how long the samples have been lingering at the CDC, though the fact that they sent them there at all would indicate it's been a while.

There's a new strain theory making the rounds. This is not your grandfather's L vs. S strain theory, but a more general look into the mutability of the virus and its effects. The authors sequenced virus samples from a small, apparently random sample of patients in Hangzhou, China, all of whom had had early contact with cases ex Wuhan. According to the preprint:

Taken together, despite only 11 patient-derived isolates being analyzed in this study, we observed abundant mutational diversity, including several founding mutations for different major clusters of viruses now circulating globally. This diverse mutational spectrum is consistent with their relatively early sampling time and relative proximity to Wuhan city, where the first viral strain was identified. The full mutational diversity of the virus in Wuhan city in the early days is still unknown to this day, due to limited sampling.

They infected a cell line with their 11 strains and observed significant variation in viral load and CPE (cytopathic effects such as cell death) between the strains. Due to the complexity of disease course they did not attempt to evaluate the pathogenicity of the strains directly in humans, but they did note that

[t]he tri-nucleotide mutation in ZJU-11 is unexpected; we note that this specific viral isolate is quite potent in our viral load and CPE assay, and its patient remained positive for an astounding period of 45 days and was only recently discharged from the hospital.

There is some discussion of the similarity of some of their samples to strains found in Washington State and Europe; in vitro, their Washington-like strains produced significantly less viral load than their Europe-like strain.

PlagueBlog complained about the weird county rate map in the state's new dashboard. Here is our less eye-bleeding version, thanks again to R:
Click to enlarge. Later today the new town-by-town numbers should be out (hopefully not in as inaccessible a format as the PDF dashboard, though R prevailed over the PDF in the end with the help of tabulizer) and mapped.

P.S. Massachusetts' numbers are out for today. We are again only up 4%, though there was a big jump in deaths: 221 new deaths for a total of 2,182. The new towns and cities data wasn't there at 4pm as promised, but it is there now:


Click (and zoom) to enlarge. Something's up in the small town of Middlefield, whose rate is about 19,000 cases per 100,000 residents (due to having 89 cases and only about 500 residents). That forced me to switch to a log scale for the case rates. Chelsea, while no longer the obvious hotspot it was, is still in the running at #3. The top 10 cities and towns by case rate are now Middlefield, Winthrop, Chelsea, Brockton, Everett, Randolph, Lynn, Lawrence, Topsfield, and Revere. Holyoke is still #11, but Littleton has dropped pretty far down the list. There are still 24 towns (in gray) with no cases.

P.P.S. Most of that news turned out to be bad data from the state. The actual results were much like last week's, with the top cities being Chelsea, Brockton, Everett, Randolph, Lynn, Lawrence, Topsfield, Revere, Holyoke, and Danvers, but it did still work better on a log scale.