Ricochet is the best place on the internet to discuss the issues of the day, either through commenting on posts or writing your own for our active and dynamic community in a fully moderated environment. In addition, the Ricochet Audio Network offers over 50 original podcasts with new episodes released every day.
Day 104: COVID-19 It’s Over, But How Do You Convince People That It’s Over?
The chart above is from the website Rt Covid-19 created by Instagram co-founder Kevin Systrom. The website purports to be tracking the effective reproduction number of the virus that causes COVID-19 by localities:
Most people are more familiar with R0. R0 is the basic reproduction number of an epidemic. It’s defined as the number of secondary infections produced by a single infection. If R0 is greater than one, the epidemic spreads quickly **. If R0 is less than one, the epidemic spreads, but limps along and disappears before everyone becomes infected. The flu has an R0 between one and two while measles sits in the high teens. While R0 is a useful measure, it is flawed in an important way: it’s static.
We’ve all witnessed that humans are adaptable. Our behavior changes, whether mandated or self-prescribed, and that changes the effective R value at any point in time. As we socially distance and isolate, R plummets. Because the value changes so rapidly, Epidemiologists have argued that the only true way to combat COVID19 is to understand and manage by Rt.
I agree, and I’d go further: we not only need to know Rt, we need to know local Rt. New York’s epidemic is vastly different than California’s and using a single number to describe them both is not useful. Knowing the local Rt allows us to manage the pandemic effectively.
States have had a variety of lockdown strategies, but there’s very little understanding of which have worked and which need to go further. Some states like California have been locked down for weeks, while others like Iowa and Nebraska continue to balk at taking action as cases rise. Being able to compare local Rts between different areas and/or watch how Rt changes in one place can help us measure how effective local policies are at slowing the spread of the virus.
Tracking Rt also lets us know when we might loosen restrictions. Any suggestion that we loosen restrictions when Rt > 1.0 is an explicit decision to let the virus proliferate. At the same time, if we are able to reduce Rt to below 1.0, and we can reduce the number of cases overall, the virus becomes manageable. Life can begin to return to ‘normal.’ But without knowing Rt we are simply flying blind.
“[I]f we are able to reduce Rt to below 1.0, and we can reduce the number of cases overall, the virus becomes manageable. Life can begin to return to ‘normal.’ ” The chart above suggests we are in a good place. Particularly when you see it has part of a trend line. The following graphs show the progress using the same calculation methodology from 4 weeks ago, to last week:
They last tweaked their model on 4/26 but re-ran the updated model against their complete dataset, so they are comparing “apples to apples.” Their model may be good or not, but it does seem to confirm observational data: the epidemic has definitely slowed in this country.
But do people really see that? Yes, a lot of people are anxious to get back to life. But are they feeling confident that doing so is the right decision? How do you persuade those that remain fearful that the “quarantine breakers” are actually common-sense individuals not just recklessly selfish?
This is going to be tough. The public has been fed a lot of data where the numbers rise steeply. As the body count mounts, and it will continue to do so, how do you get people to realize that things are winding down? Observational data is always the most powerful: “Who are you going to believe? Me, or your lying eyes?” the old joke goes.
The President has shifted his focus away from the health crisis to how we get our economy back. That is a start. The people going to the beach and surrounding their capitols and city halls also contribute to the pressure to get back to work. But media and the progressives are highly invested in the narrative of fear. They will continue to amplify “confirmed cases” and obscure the fact that the rising case count is actually good news so long as the rate of serious illness and death continues to fall. Testing confirms the prevalence but not the severity of this illness.
Keep an eye on hospitalizations in your area. If hospitalizations continue to fall, it is really over. Ask your social and church groups when they think it will be ok to meet in real life? Prompt their questions about how to interpret what their eyes are seeing.
One by one, we emerge into the light.
[Note: Links to all my COVID-19 posts can be found here.]
Published in General
Rodin, I don’t think that it is over. I think that as soon as we loosen restrictions, the spread will increase, and then deaths will increase. I think that we need to face this.
I am very doubtful that Rt is less than 1.
How in the world are they calculating Rt, when we don’t have good data on the number of infections? I don’t see how this is possible. It may be wrong, of course, but I would take a fair amount of convincing. For this reason, I am highly skeptical of their charts.
Of course, the worst part is in the comment that you quoted: “Any suggestion that we loosen restrictions when Rt > 1.0 is an explicit decision to let the virus proliferate.” Yep, and “people will die.” And as soon as we ease restrictions, Rt will increase again, and we’ll have to lock down again, apparently indefinitely.
Am I missing something?
I would guess that the % if pretty significant…a lot of people *really have* been sheltering in place. And in low-population-density areas, there is less opportunity for transmission. The % unexposed must vary a lot from area to area, but for the US as a whole, I bet it’s nontrivial.
I located the classical mathematical model for epidemics, the SiR model, (susceptible–infected–removed/recovered) which was first developed in 1927, was apparently forgotten and revived in the 1970s. Here is a very nicely done calculator which allows you to modify the assumptions and see what happens in the simulated environment:
https://gabgoh.github.io/COVID/index.html
This implementation of the model allows you to implement “intervention” (aka lockdown) on any specified date; unfortunately, it doesn’t allow you to *remove* lockdown at some later time. And per my earlier point, it treats the population as homogeneous, whereas IMO the correct approach would be to segment it into geographies with differing behavior but with some travel between the geographies and consequent virus transmission.
While Arizona appears to now be in the “green” Coconino County where I live just had its highest number of cases in the week that ended yesterday.
Perhaps it would make sense for decisions to be made county by county in many states. For example, clearly there need to be stronger measures in the Detroit metro area, however does this mean that the UP needs to be on lockdown? (Paging Arafant.) New York City has one fifth of the number of cases, but should the rural counties in upstate New York be on the same restrictions?
We don’t know. And some voices, such as those that are hating on Sweden, seem to be terrified that we may find out.
The wife and I went to a restaurant for lunch today. We were the only patrons in the dining room. Service was great and the food was served up very quickly. They had half the booths shut down and most of the tables stacked in a smaller side room. I think a few other folks were in the bar half of the building So maybe 4 or 6 folks in an eatery that could sit 150.
I am using the term “over” in the sense that we now know the outcome and that it isn’t going to be an existential threat — certainly not one that justifies the extensive interference with liberty. Yes, there will be additional deaths and illness, not just growth that was in the past and that panicked us. I don’t pretend to understand the Rt model works but as stated in the OP the trend in the data seems to make sense.
@garyrobbins, ignore case counts and focus on hospitalizations. According to the Coconino County COVID-19 website, you currently have 20 hospitalized COVID-19 patients, down from 81. Although it is possible that some COVID-19 victims die without hospitalization, I think to the first order tracking hospitalization counts gives you a realistic picture of how the epidemic is going even though there will be additional cases and deaths. Remember increased testing is identifying increased cases but doe not mean that the number of seriously ill is going up at the same rate.
The co-founders of Instagram have been working on a project to collect & display Coronavirus information, including state-by-state estimates of Rt.
https://rt.live/
They also provide a link to another site that takes a different approach to calculating Rt.
can high ceilings make a difference?
Thank you. I feel much better.
What do you think of the analysis at 538? See https://projects.fivethirtyeight.com/covid-forecasts/?ex_cid=rrpromo
Remember, most people who test positive have zero symptoms… Based on the Diamond Princess data, the younger cohort, 20s 30s 40s are more likely to display mild symptoms.
2 to 3 percent of postive cases are ‘serious’… 97% are mild or zero
80-90 percent of the population will be negative despite exposure.
All 6 models essentially tell the same story.
I would be shocked if the death count exceeds 99k
If I were a betting man (and I am), I would take the under (100k)
Northeastern and Texas seem most credible to me.
These are my assumptions:
I agree with Aaron Miller that your assertions are somewhat simplified. Yes we should open up everywhere soon, even in NYC, but we don’t really know what will happen. Infection and hospital utilization curving down with people staying home doesn’t predict the extent to which the virus will return when they’re out and about. Maybe we now have our Medical ducks in a row, but that’s all staying home may have accomplished.
It depends entirely on the question you are trying to answer. I can think of a number of reasons it wouldn’t matter, especially if all you are interested in is whether enough medical resources exist at points in time. In that case, the number of hospitalizations per day is all that matters. For simplicity, assume that every case of COVID-19 results in a hospitalization, then build a model based on the observed number of hospitalizations and R(t) and use the data to estimate the number of hospitalizations. It avoids having to know or estimate a lot of other parameters–but it also a model that is completely useless for estimating the level of infections needed to build herd immunity.
That has been the biggest problem with policy modeling for this pandemic. We have a lot of institutions with a “Have model, will travel,” mentality. There are several classes of experts that need to be consulted for these policy decisions, each has models built to explain the phenomena that they study. The problem is that each group’s models are only useful for the questions that they are interested in, and for the most part the decision makers seem to be relying on the same model to provide insight on phenomena they were never designed to explain.
Similarly, if you are comparing states to see what actions/conditions were most effective to flatten the curve, you would want to make sure that you are measuring the same thing in the same way in each state. Otherwise you have less confidence of the impact of different policies. For example, the testing in my state appears to be concentrated around essential industries. As I understand it the employer screens every day for fevers, and if someone has a fever, they test. This is not how scarce testing resources are rationed everywhere else. So, it seems likely that this procedure picks up COVID-19 cases that would be missed everywhere else, but it also has the potential to skew the number of infections in the whole state relative to other states.
So short answer is “it depends.”
If you ignore NY and look at 49 states the result is:
death per million = 141
‘typical’ flu season = 100 death per million
minor digression: Trump lost the popular vote in 2016 but if you ignore California, he won the popular vote for the other 49 states
In Coconino County, Arizona we have 300 deaths per million.
Thank you for the link. Although I intend to study it more my initial reaction is that it understandably focused on a non-contextualized metric: projected deaths. As much as we don’t want people to die, people are going to die, and a lot of people are going to die of a lot of different things. The question is how do you use the data to inform public health decisions? And that requires context. Just like comparing auto deaths to airplane crashes, you look at distance traveled. So, too, you can’t assess the significance of COVID-19 deaths and the public health reactions without comparing it to other diseases. And you can’t make those comparisons until you know enough about the disease to know the probable infectiousness and severity. We seem to have established that the number of infections is many multiples of confirmed cases. And confirmed cases are multiples of hospitalizations (which in my opinion is the best metric for the level of seriousness of the disease). Deaths are only a fraction of hospitalizations. The infection fatality rate (IFR) is probably 0.03-0.05% compared to an IFR for seasonal flu of 0.01%. It is more deadly but nowhere near the deadliness of the Spanish flu of 1918-1920, with an IFR of 10%. When the virus first appeared and based on the reports out of China and what seemed to be happening in Iran and Italy, the shock turned to panic. But as we have gotten our own domestic data and information from Germany and Great Britain it has become obvious that this is not the threat we feared. [I am writing this from memory without going back to sources, so my IFR statements regarding COVID-19 and flu could be off, but scales.]
I agree that the state is far too large to be useful. I have been thinking about how one can use the available data to gauge their own risk.
This display (https://covid19.biglocalnews.org/county-maps/index.html#/county/48043) will give you county data and every state has wide variations by county. Try it with your own county. My link should go to the largest county in Texas, but one that has a very small population count.
But can you draw risk conclusions from the county data? Unfortunately, no. The county I live in has wide variations over its area. The county health authority has provided data by zip code (there are about 8 zip codes in the small county I live in). Does that help me detect my personal risk?
No, it does not. The zip code I live in has the majority of cases and the majority of deaths in our county. But that is because one nursing home is responsible for many of the cases and (if memory serves) half the deaths for the entire county. So maybe the right risk number should be considered for an average person not in a nursing home. But a big family living in one home could all get exposed by one of the family members, and they would then become a mini hot spot.
Your personal risk is complicated. Geographic statistics do not tell us much about that.
At my age, acquiring this infection would be a very serious event. On the other hand I’m retired so maintaining social distance doesn’t cost me anything financially so I can plan to be very cautious. Somebody who really needs to get back to work has to think about it more.
Masks are not a bad thing. Incarceration and social isolation are bad things, but masks are not. If we had passed out masks in February 2020 we could have kept society open, including the opera houses and the hockey games. Mrs Doc Robert and I would be heading to Amsterdam on Wednesday for the Mahler Festival instead of continuing to live under quasi-house arrest.
Wasn’t someone telling us back then that masks were a bad idea and would actually cause disease?
The doctors telling us that were lying, so that the hoi polloi would not be competing with them for the available masks.
If I had to guess, Navajo nation has high death rate because of co-morbidities associated with alcoholism
Having our medical ducks in a row was all staying at home was supposed to accomplish. Flatten the curve so the hospitals won’t be overwhelmed. We did and they weren’t.
Oregon’s hospitalizations hit a new low, down 40%.
Sometimes it’s necessary for the government to lie to the people; otherwise the people might lose their trust in the government.
That’s all staying home was *supposed* to accomplish.
We do know what it looks like without restrictions. It looks like NewYork in March. The Ro is intrinsic to the virus. Best estimates are that COVD19 has an Ro of 2-3.5. The disease remains highly infectious. We can hope warming weather can drive Rt down which will allow social distancing to ease. But right now Rt has dropped to less than 1 because of the extreme measures. Nothing about the virus has changed. It is important to remember this. We have changed our behavior. Therfore changing it again will alter Rt again. I doubt we could be more distant therefore this low represents the minimum of spread. It can only go up as we roll back distancing practices. We need to at least keep Rt to 1. We also need to understand that right now we are in a place with many more infected people from which the virus can spread than we had back in February. Even going slightly above Rt 1 will see a nominally large spike in new case. If it jumps back to 2 or higher it will be a second wave. Also keep in mind this disease has a fairly long lag time between infection and symptom. You might not see the second wave until it’s been building for two weeks. Which is where more testing comes in. We still dont test enough to see a second wave building until people are showing up to hospitals presenting symptoms.
the mask isn’t about you- it is to help others. Most of life isn’t about you either
This is a wonderful story that just popped up on my Yahoo! news feed:
The residents and staff remained well for the 47 days that the virus was hitting nursing homes hard in the region all around them.
What is so interesting about this is that it confirms the storm-like nature of the virus as it becomes active in a particular place. Looking at the staff and patients emerge after the isolation period confirms the weather part of the virus equation. As the weather moved out of the 40 to 52 degree band and became more humid, it simply went away. It is exactly the way flu operates. As soon as the weather changes, it simply goes away.
This article explains this phenomenon. [The article dates back to 2014, so it’s reliable. (Isn’t it sad that I cannot trust the objectivity of a single thing I read right now about covid-19 because it has become political.)]:
There’s no question, looking at those pictures from the nursing home in France, that this storm has passed. There will be another snow storm next winter, but for now, it really is safe to go out and to be together.
That knowledge–it is a flu-like contagion–creates a very rational decision-making framework for the public.
I keep thinking of all the catch phrases that my parents and grandparents would say whenever we kids were trying to get through tough moments. “You made it!” “You’re in the clear!” “We’re out of the woods now.” “The storm has passed.” There is a wonderful recognition of this moment in the pictures of the nursing home staff and patients in France in that news story. :-) Obviously the little people in the world–not our wise and highly educated and highly paid politicians and leaders–know something in their gut about the passing of this year’s covid-19 attack. They knew they could stay in with the residents for a while because they knew the virus storm would pass.
There are so many lessons for leaders in this remarkable story.