but would cost even many more lifes…
What’s the math for that?
This regional / ICU-capacity approach makes way more sense to me than what the state.has done to date.
1 Like
It’s a move in the right direction, but too late and too harsh.
Using occupied ICU beds as a basis for the strength of restrictions, and regionally-applied mandates to reduce “mixing”, are both good ideas. (Not all patients in the ICUs are there due to covid, but that’s a secondary point.)
I presume that the projection that So. Cal. will reach the 85% occupied ICU threshold in a few days is based on an extrapolation of the slope of the ICU beds occupied vs. time. It’s that slope that should be used as the feedback control. Cutbacks and relaxation in the amount of allowed business activity should be based on the amount of departure of the actual slope from a safe slope – not the number or percentage of ICU beds in use itself.
Knocking back business occupancy from 50% to 20% capacity if we’re even the tiniest epsilon over 85% ICU capacity – and not doing so if we’re an epsilon under – is preposterous.
Instead, when the slope of the number (or percent) of ICU beds occupied vs. time curve exceeds some critical value that, if maintained for some specified minimum length of time, like a month, would lead to 85% ICU beds in use, allowable business capacity would drop from 50% to 40%, say. If the slope remains the same or goes up further, allowed occupancy would drop back further, from 40% to 25%, and so forth. Conversely, as the slope of the curve goes down, the allowed occupancy level would rise in the same manner, gradually approaching 100% as full control (i.e., the slope we’d have in the absence of covid) is reached. Which may not be until late next summer
The idea is to have at least a month of projected ICU capacity available at all times. Not wait until three days before an extrapolation shows that 85% of beds will be occupied, and then slam on the “emergency brake”.
Far less harsh, true feedback control, and it would have kicked in earlier.
Also, I don’t see the rationale for business closures and cutbacks across the board. What fraction of covid outbreaks have been traced to hair salons, for example? And how many have been traced to grocery stores? How many to outdoor dining other than in makeshift, essentially enclosed “tents” on the street?
After being fairly stable for a few months, since the beginning of November LA County’s ICU occupancy has increased by 421, from 223 to 644. At that rate, the 486 remaining ICU beds would be exhausted by mid-January.
Given that exponential rate of increase, it’s not too harsh, and given how sudden and recent the change was, it’s not very late. Nobody knows exactly what caused it, so nobody knows how to fine-tune a shutdown to target the problems.
That’s quite the dramatic plot.
Looking closely at the timeframe between about mid-October until now, the curvature (second derivative) of an eyeball curvefit appears to me to be negative; zero at most. Meaning that the slope of ICU beds occupied vs. time has been increasing since then.
I guess my point here is that control should have started in mid-October, with ongoing, smaller adjustments of the "dimmer switch – both ways as warranted, and much more frequently.
ICU bed occupancy is something that’s related to latency of infection, true, but it’s a hard number that’s not vulnerable to debate and accordingly a good “anchor” on which to base metrics.
Oh, really? Easy to say with 20/20 hindsight.
Yes, really. That’s the sense of “latency” I’m talking about, which I agree is problematic but have no doubt is understood by experts in control theory.
Back in September, a mathematician would have detected a rapid change in the second derivative of a curvefit to the extant data up to that time. An upward-trending slope would have been predicted, so that even possibly as early as mid-September, restrictions initially tightened. Gently at first, maybe just by a few percent.
The second derivative should also play a role in a more sophisticated feedback loop, in other words – I would think.
Using ICUs as a measurement for anything is completely wrong or as a an epidemiologist recently said - using ICU is like using a fire detector when the house has already burned down - it’s way too late as an indicator.
Yes, the ICU load today essentially reflects what was initiated maybe two weeks ago.
But what else would you use instead ?
Honk, I’m not myself a control theory expert. Far from it. I had one course as an undergrad. But I’m pretty sure that a control theory specialist would tell you that quite a bit can be gleaned from the combination of prior and current data.
I’ve always thought that deaths, not cases, was the overarching concern. Deaths and ICU occupancy are closely correlated, at least in SD County.
I agree with Newsom on this one, using hard ICU data.
Sorry, but every time somebody points to huge flaws on your suggestions you avoid any specific answers by saying you are not an expert. At the same time every time somebody points out to you that experts say that your “fine tune” method doesn’t work you just say that you think it has to work. Either you come up with serious suggestions and not some half-baked, obvious flawed dreams without any excuses or you should listen to experts who know what they are talking about (As said many times before there is no measurement available which would allow any “fine tuning” - as you now said ICU are useless, same are general hospitalization. The only possible way would be infections but as said before many times that would require a much, much higher number of daily available test - as experts have often said somewhere between 15-20 million/day would make it possible). So stop complaining about “fine tuning” if you can’t provide any serious suggestions beside “control theory specialist would tell you…” when those couldn’t also do anything without any useful data
1 Like
OK, I hear what you’re saying Honk. I really don’t agree with you, but that’s OK. If I recall, this isn’t a scientific journal, it’s a food talk forum. I get to express my opinions, and you yours. Let’s not get into a pissing match.
That’s not a pissing match but just a discussion about logic and reasoning - one can’t complain about something without providing any solutions and otherwise hiding behind the argument that others are the experts and should know.
I’m not hiding. I like it under the bed.
I’m wishing that a control expert would speak up! I’m always willing to stand corrected.
Really, I basically agree with you. It’s a topic worth discussing – back and forth. And it does have more than a passing impact on food.
I think the exponential upturn in the curves in the November (which you can see in cases, positivity rates, and deaths as well as ICU) are due to the change in the weather. At the places where people were already getting infected—at home, at work, on public transportation, dining indoors, gyms, etc.—they turned on the heat and closed windows, lowering the humidity. That doubled or tripled the concentration of virus particles, thus increasing the R0.
So I don’t think there’s anything in the data before late October to predict that that was going to happen. Some experts thought it might, but we didn’t know, and it’s still not proven…
But that’s the point - a control expert doesn’t know anything about virus and epidemiology - the expert you need is an expert who tells you which data is useful for this particular virus to even think about any “fine tuning” (And over the last several months I have set in a quite a number of talks and discussions from those and it is clear what their opinion is regarding data and what is possible (or better not possible) with it (and all of them said that ICU data is useless to control the pandemic - you need infection data, which we don’t have on that level and so no control expert could do anything)
1 Like
Interesting direction of thought. We disagree.
Frankly, I think that by relying only on epidemiologists, etc., as we have, we’ve lost a lot of time. And lives.
The curvature of this plot first became non-zero positive around Sep 1. That’s very clear by inspection.
From Sep 1 to mid November, the curvature (second derivative) was almost constant,
When the second derivative initially changed from zero to a non-trivial positive value in early September, action could and should have been taken, with guidance by way of any half-way sophisticated algorithm.
In mid-November the curvature seems to have gone back to near zero, but this time with a steep positive slope (first derivative).
And now, due the delay to act after a clear problem was visible in the data in early September, the main breaker of The Entire Economy in our usual regional paradise may be flipped to the Off position completely, or almost so, again.
It’ll happen abruptly and bigfooted as usual:
In just a few days – and with little notice – the latest heuristic seat-of-the-pants intuitive autocratic decree from Sacramento will start to compel our behavior again, like children. So.Cal. businesses will abide and suffer, but people will party on.
There has to be a softer, more effective control than cancelling restaurants, salons, gyms, etc.