By Jeff Hester on Apr 21, 2020 11:51 pm
You know those nice charts and graphs that make it look like we are over the hump of COVID-19 and that things are about to get better? I hate to break it to you, but those predictions are dead wrong, with an unfortunate emphasis on “dead.”
Over the last few weeks there’s a pretty good chance that you’ve heard that the worst of COVID-19 is behind us. It would be lovely, but is it true? To answer that question many people turn to the predictions of influential and oft-cited computer modeling by the Institute for Health Metrics and Evaluation at the University of Washington. If you take those models at face value, it might look like deaths are about to start dropping fast. By June hardly anybody will be sick. Let’s celebrate! It’s all downhill from here.
There is a fly in that ointment, though. None of it is true. Instead of dropping rapidly like the models predict, the data show that they won’t, and we have every reason to imagine they are going to start going up again.
The fancy models you’ve seen on TV aren’t really epidemiological models at all.
Before talking about how I know the models are starting to fail let’s take a look at why that was bound to happen. To really understand the IHME models you have to look past the popular media and read their technical papers. The most important thing to understand about the IHME “model” is that it’s not an epidemiological model at all, it’s just a big “curve fitting” code. In other words, its a fancy version of looking at a bunch of points on a piece of paper and drawing a line through them without really thinking too much about what that line means.
Specifically, the group at IHME take the shape of the curve from Wuhan, China and, with a few adjustments, run it through data from the United States. Or rather they fit the early part of that curve to data from the U.S. Then they say, “If the United States behaves the same way that Wuhan did, then this is what the future ought to look like.”
By the way, I’m not trying to pick on the IHME group. Their original goal was not to predict the overall course of the pandemic in the United States. They just wanted some idea of how high the crest of the first wave might be to help health care organizations predict what they might be up against. They are completely up front about the nature of their calculations, and they present wide error bars. People talk about a rosy picture where only 60,000 people die, but the quoted uncertainty of the projection extends to over 140,000.
Anyway, back to the issue at hand. Those nice plots you’ve seen depend on the assumption that things in the United States will go more of less like they did in Wuhan, China. Ummm… Whoops?
The United States is NOT Wuhan.
Right out of the gate, it’s not too surprising that the IHME models have done reasonably well up until now. We have been on the rising side of the first wave of the COVID-19 pandemic. The early part of that curve reflects the time before any actions had been taken. It just shows the original unchecked rise the disease in a dense, urban setting. New York City, the first major outbreak in the United States, is a lot like Wuhan. It makes sense that the disease behaved in similar fashion.
As things started getting bad in Wuhan the city responded with serious lockdowns. Public transportation was shut down. Schools and businesses were closed. Factories were closed. No one was allowed in or out. The streets were empty. You stayed in your house or apartment. These restrictions were enforced.
Here in the U.S. the first cities and states that saw many COVID-19 cases (e.g., Washington, California, and New York) were quick to respond with what remain some of the most effective lockdowns to date. Those lockdowns have been largely respected by the populace. Those places are as close to Wuhan analogs as we have here. Unsurprisingly, as long as those states dominated the statistics for the nation as a whole, the IHME predictions did reasonably well.
It’s time to expect the COVID-19 model predictions to start failing.
That is where the similarities between the United States and Wuhan end. Even the tightest restrictions in the United States are like a day at Disneyland in comparison with the two month lockdown in Wuhan. Even the restrictions that remain in place after Wuhan “opened up” are tighter than what most Americans are experiencing right now.
Further, as the number of cases in the United States increases, states that were slow to act or that have few if any restrictions are starting to account for more and more of the overall statistics. Some states, with encouragement from the Rose Garden, are even remove what restrictions they do have. Compare that with the IHME models which assume that restrictions here will be as effective as those in Wuhan, and once imposed will remain in place indefinitely.
Figure 1: A comparison of the COVID-19 death rate (gray curve and data bars) with the COVID-19 new case rate from a week earlier (red curve and data bars). The heights of the two curves have been shifted to match at the peak. Note that the two curves match each other exactly. Using these relationship we can use the new case rate over the past week to predict the death rate in the coming week. Based on such an extrapolation it looks like the death rate over the coming week should remain fairly flat. (Data compiled by The New York Times.)
In short, on the front side of the current COVID-19 outbreak things in Wuhan were a reasonable guide to what might happen here. That’s no longer true. Of course the IHME models are going to start failing. They use assumptions that simply aren’t true. As they say back in the neck of the woods where I grew up, that dog don’t hunt.
Even so, if you glance at the IHME model page today things don’t look so bad. The model does a pretty good job of predicting death rates up to this point, so why am I claiming that we have reached their breaking point? To tackle that one, we need to look a bit more closely at the data.
If you know how many people are sick this week, you know how many people will die next week. It’s more than you think.
Death rates are known as a “lagging indicator” of a disease outbreak. They are what come last. Before people die, they get sick. If you know how many people are sick today, you ought to be able to predict how many people are going to die in the future.
This way of predicting future deaths works amazingly well with COVID-19. Look at Figure 1. The gray bars and the thin gray line show the COVID-19 death rate through April 20. The red bars and the thin red line show the rate of new cases, but shifted forward in time by one week.
Figure 2: The red line on the right side of the figure is a prediction of the COVID-19 death rate over the coming week based on the new case rate from the previous week (see Figure 1). The dotted blue curve shows the projections of the April 15 version of the best-fit IHME model. Instead of falling, as predicted by the IHME model, the death rate will remain fairly flat over the coming week. For reasons discussed in the text we have reached the point that deaths can be expected to greatly exceed the IMHE model predictions. This has profound implications for the future of the nation. These models are being used to support a false narrative that we are over the hump and that we can start opening back up. (Projections from IMHE website as presented on April 21, 2020.)
If you can’t tell the difference between the two curves, that’s because they are indistinguishable. That’s the whole point. Since the pandemic began, knowing the number of new cases today has allowed you to predict the number of deaths a week from now with extraordinary accuracy. When you do that right now, you find that going forward (the red line and data bars on the right side of the figure) the death rate will fall only slightly, if at all.
Now that we know what death rates are going to look like for the next week, how does that compare with what the models predict? Figure 2 gives the answer. The figure just shows the same data from Figure 1, but now overlaid with the April 15 version of the IHME best-fit model. The solid blue line shows the actual death rate up until this point. The dotted blue line shows what the model is predicting. Notice anything?
A week from now the death rate is likely to be close to double what the models are currently predicting. The numbers of deaths are going to remain flat. And after that, they are going to start going up again as death rates in states without lockdowns rise.
Mass murder by any other name remains as horrific.
There you have it. Things are not about to get better. We are not going to hold deaths to anything like 60,000. Death rates are not going to start falling. Quite the contrary, as states without serious restrictions start to account for more and more of the total, death rates are going to start climbing again.
So the next time you hear someone say that the worst is behind us and that it’s safe to start opening up again, spit in their face. Kick them in the knee. Slap them around a bit. Knock some sense into them! They are wrong. And if they have their way, lots, and lots, and lots of people are going to die needlessly.
Let me say that more accurately. Lots, and lots, and lots more people are going to die needlessly. Because had we listened to epidemiologists in the first place, today the United States might be looking at few hundred deaths instead of 45,000 and climbing.
When I think about public officials who fail to do what they can to lock things down, or who talk about reopening soon, the term that comes to mind is “mass murderer.” The term may not be polite or comfortable, but it is not hyperbole. Because officials who fail to lock things down when they could have are killing people as surely as if they held a gun to their heads and pulled the trigger. Sorry, but “mass murderer” is the correct term.
Futures on pine boxes are looking good.
I keep wanting to be wrong about COVID-19. I really wish that I were being alarmist, but so far I’m batting 1.000. I’m confident enough that things are going to start looking worse than the IMHE models that I’d probably bet a good bottle of Scotch on it.
So what does the future hold? That’s an interesting question. But if you are looking for investment opportunities with a growth future, I’d probably go with coffin makers, mortuaries, and crematoriums. For them, things are looking up.
The post Pine Boxes ^ Invest now, the numbers are going up appeared first on Dr. Jeff Hester.
Read in browser »
    
Recent Articles:
|
