To recap, though, and put some things in perspective, one can look at the plots in the SD Union Local Section of the rolling 14-day average of positive tests (“positivity rate”).
That number was dropping rapidly until May 20, when restaurants were abruptly allowed to fully open, howbeit with lots and lots of rules. Shortly thereafter, the positivity rate stopped dropping. It leveled off at 3%, where it remained for about two weeks, which appeared to be stable at least. Bars were then abruptly allowed to fully reopen (also subject to many rules) on June 12.
On June 19, the positivity rate briefly dropped to 2.5%. Unfortunately, after that encouraging minimum, the positivity rate began to rapidly and steadily go up. Bars were abruptly closed on June 29. The positivity rate continued to climb. Restaurants were abruptly closed on July 6. The positivity rate is now around 6%.
If the State and/or County gets a headache when thinking about employing a stability and control expert, here’s a simple modification to the existing On/Off approach:
Three days after the June 19 minimum in the rolling 14-day average positivity rate, on June 22, it was clearly rising rapidly. In my view that’s when a correction should have been exercised. Not waiting for 10 to 17 days.
But the correction wouldn’t have been the abrupt and complete closure of bars and then restaurants. It would have been a scale-back of some sort of both, on June 22, such as reducing the number of opening days; “dimmer switch” control, if you will, as opposed to On/Off. If the initial scale-back worked, the correction would be held until the positivity rate was leveling off again or declining. If not, there would be further incremental scale-back. If so, the opening days could be incrementally notched back up. And so on.
I don’t really understand why you think any kind of “feedback control” should be part of the calculation for viral spread. I had, job-related, a number of meetings where well known epidemiologists and virologists explained their models and their theories how to counter the epidemic (from a practical and theoretical standpoint) and none of them had any thoughts on feedback control. No offense, but just because you had a few lectures in feedback control for stability in engineering some years ago doesn’t mean that the same concept has any/significant value in the world of epidemiology.
But if it didn’t work initially and you need to continue scaling more back you will have killed many additional, completely unnecessary people during this time (and it will take about 2-3 weeks to see effects in hospitalization rates (which is (beside death rate))a key indicator of success and to a much kesser extend the positive test rate). We can’t afford to kill potentially more people if we can avoid it with a complete closure.
Our objectives overlap: Reduce deaths to the absolute minimum. Zero for all time is impossible, but a very tiny fraction is. We’re talking about the best way to get there.
I say bring in the experts in feedback control and let them weigh in on this. That’s all. My opinion is that it does have value in the area of epidemiology. And I’m a little surprised that no one has even talked about it in your meetings, because feedback control is being used now.
There are people talking about it in such meetings but they don’t see great value in it in the current situation. In addition, with so many easy unforced errors already made in controlling the pandemic, adding such “fuzzy logic” approach would be way too risky and unknown to kill even more people - until now you still haven’t explained how you want to avoid even a single unnecessary death with your approach when you even admit that it might be necessary to further “notch up or down” during the process. Wouldn’t you agree that this would be too risky already ?
I think that we’ve waited way too long to exert control, and that by the time we’ve taken action, there have been far too many new infections. “Notching down” wouldn’t exclude complete closures over the same (long) time period that “full open” was allowed. It would start reducing opening days much earlier, thereby reducing transmission earlier and saving lives that were lost due to the long delay before a sudden “full stop”. Re-opening would be similarly ramped upwards a notch at a time, based on feedback. “Full open” would be a goal. Just not instantaneous either way. So I see it as less risky, not more.
I would implement a nation-wide lockdown for 1-2 months (similar as it was done in many parts of Europe) to get the numbers to a similar level as in those countries. They will also get some setbacks in the future but it is obvious what really works are nation-wide lockdowns (which are actually enforced) to crush the curve.
Completely agree but without a national lockdown there won’t be any next steps which would have a chance to be successful. Any future plans can only work if we get the infection, hospitalization and deaths rates drastically down nationwide now (before in 2-3 months the flu season will start and the problems will really start)
There are no definite conclusions drawn in this article, but it’s certainly a worthwhile read in the context of the discussion here.
You and Robert are both probably right, though. In the absence of a national strategy that’s widely accepted by the public, “asking” doesn’t work very well here.
On the subject of engineering-level feedback control as it applies to covid, here’s a very interesting and relevant article from the highly respected IEEE.
That article seems to me to describe pretty much what California has been doing. They track multiple indicators for each county and have five stages from lockdown to fully open.
I think you’re fantasizing about a level of granularity that’s impractical given the many unknowns about the disease, inadequate testing and contact tracing, and rushing on Newsom’s part. No county should have been allowed to go from one stage to another without a month of watching in between. And the practicality of having different counties on different timetables is pretty dubious given the state’s inability to limit travel by irresponsible citizens.
Control theory is not unfamiliar to epidemiologists.
The whole point of a rolling 14-day average is to smooth out granularity in the data. When that rolling average shows a distinctly rising trend, as it did after June 19 in SD County, it’s not at all a granular change. But I agree that working out a practical way to modulate opening days or hours would need careful thought.
