[leourbina]

This style of analysis has the implicit assumption that the timeline of confirmed cases is accurate. Despite the infection growing visbly rapidly in NY, it could be argued that increased testing in NY state has been a contributing factor to the perceived acceleration of the rate confirmed cases compared to other states with fewer tests. To this day throughout the US there are many reports of people who present symptoms at home and are not able to get tested. Without testing uniformly throughout the country, how can we make any kind of comparative analysis of how different state's infections are progressing?

How are any of these analysis taking testing rates into account to truly understand the spread of covid19?

[olliej]

I mean you can also use actual fatality rates as a proxy for the actual infection rate, that's harder to artificially inflate or deflate, and CA is much below NY there as well.

[pmiller2]

Fatality rates are still subject to bias due to lack of postmortem testing if it's not a confirmed case. This would likely include almost everyone who dies outside of a hospital.

[wahern]

For the past couple of weeks I've been plotting the data from covidtracking.com, along with generated data for perfect 3-, 5-, and 7-day doubling. On a log scale the graphs unequivocally show California has been on a 5-day doubling curve since almost the very beginning. NY was on a 3-day curve until a couple of days ago. The US curve follows NY, which isn't surprising. Unless testing criteria changed day-to-day in a very specific, mind blowing coincidence, the difference in infection rates is pretty clear.

And in case you haven't played with the data yourself, the difference between a 3- and 5-day curve is huge in terms of the absolute numbers you reach. As the article says, faltering for just a few days absolutely could have made all the difference between the infection getting out of control. Of course, NYC is far more dense than the Bay Area and so it's no surprise how quickly NYC got out of control. But 1) SF could have gone in a similar direction, but didn't; and 2) plenty of other regions with similar or even less density to the Bay Area were or still are on a 3-day curve.

The article also fails to mention that Bay Area tech companies started work-from-home a week before the Bay Area shelter-in-place order. That may have helped considering that the inflection point for positive cases in California (on a non-log scale where the curve starts to shoot upward) occurs at about the same time the Bay Area ordered shelter-in-place, but there would have been a lag between infections and hospital visits. OTOH, the "inflection point" may be meaningless, and in any event the numbers were rather small at that point.

The trends are extremely clear. Quibbling over testing criteria is pointless. Though, if we ever want to get past sheltering we absolutely need massive, randomized testing. And unfortunately even living in the Bay Area I'm not seeing much if any activity or news on that front, even though we should already be well along that path, at least in terms of planning and organization. All the talk about tracing contacts is a side-show; tracing is neither necessary nor sufficient. In the absence of tracing you can simply institute sheltering orders in hot spots. But without widespread, pervasive testing you'll never find those hot spots quickly enough to squash, tracing or no tracing.

[donavanm]

And in those cases you look at cause of death. Are pneumonia or respiratory related deaths elevated compared to last year? This is exactly the type of correlation that was being used to estimate infection in places like africa without available testing.