The Allure of Simplicity

It is enormously tiring to keep track of so many states and manage so much data. In the last week, the COVID Tracking project (which is my preferred COVID data source) added even more metrics. They are now tracking tests (positive, negative, and pending), antibody tests, current and cumulative hospitalizations, ICU beds, and ventilator use, recoveries, deaths (confirmed and probable), and positive cases (one case = one person but one person might have taken a test multiple times).

pictured: your average epidemiologist working with his data

Not every state reports on all these metrics and not all states report the metrics the same way. There is lag in every one of these metrics (including the ongoing significant lags over the weekends). No metric is a fully complete and accurate view of the current situation. We continue to plow forward in a blur as reality consistently comes into clear focus about two weeks after we had to live through it.

I say all this as a intro to today’s issue because there is currently a lot of uncertainty around the existing data. We’re seeing new patterns and it’s entirely appropriate to wonder what is going on and be skeptical of simple, clean explanations.

In this issue:

• Beware Simple Answers

• Why Deaths Is The Gold Standard Metric

• COVID And The Protests

• Disney Shorts: The Three Blind Mouseketeers

Beware Simple Answers

Here’s a fun story. (Fact check: Are stories still fun? I can’t tell anymore. Anyway, you’re stuck with this one.)

In my discussion on Friday about surges (subscriber only) I looked at case increases, the % of positive tests, and deaths as a set of metrics that work together to give us an overall sense of what is happening with a COVID outbreak.

Of course, the metric that I let slide through the cracks was hospitalizations. So naturally that is the only thing people want to talk about these days. I’ve now heard several times that hospitalizations are the most important metric that we should be paying attention to.

Forgive me for recycling content but I’m going to borrow from my own twitter

PoliMath @politicalmath

Also, be super skeptical of people who bounce around with their metrics. If they're talking about case increases one day, then acting like hospitalizations matter the next, then freaking out about ICU beds 2 days later, they probably don't have a solid grounding in what matters.

Yes, hospitalizations matter. ICU beds matter. But what really matters is understanding the pros and cons of each metric and being able to balance a sudden surge in one metric against a lack of surge in another.

The one metric I’m laser-focused on is deaths (for reasons I will explain in the next section). I have been watching deaths extremely closely as my primary indicator of how bad things are. As I’ve been watching these last few weeks, we have seen a surge in certain regions in both cases and hospitalizations. Both these things have possible proxy explanations.

Cases could be surging due to increases in testing. That is the most likely and also the simplest explanation.

Hospitalizations could be increasing because hospitals are opening back up. Let me explain this theory: Many hospitals have been at a severely reduced capacity for months to a point where they are losing billions of dollars. People are missing cancer treatments, dialysis, they’re ignoring the signs of heart attack and stroke, they’re postponing elective (but necessary) surgeries, they’re self-caring for minor-to-medium injuries so that they don’t have to go to the hospital.

As we’ve inched passed the first wave, people are returning to the hospitals for regular procedures and, as part of that process, they are being tested for COVID. If they test positive for COVID, even if they are there for a hip replacement, they count as a COVID hospitalization. This seems to make sense.

So that’s my theory! It is that the surge in COVID hospitalizations is due to hospitals re-opening and testing people and people with COVID who came in for non-COVID reasons count as COVID hospitalizations. It’s a coherent story with cause-and-effect, it fits in with other things that I Know To Be True™.

And then a friend sent me a screenshot of her hospital’s COVID metric dashboard. (No, I’m not sharing it, I’m not a snitch)

Here’s the reality: I was sort of right, but also not really because things are really complicated.

It looks like, in this particular metro that my friend is in, in her particular hospital, over the last two weeks, there has been *some* surge in COVID hospitalizations that is just related to the hospital getting back to normal business. My theory accounts for maybe 40-60% of the hospitalization surge.

It looks like *maybe* on top of that there is an “organic” surge of new hospitalizations. Or maybe those people are simply more comfortable coming in for care now that it seems less dangerous?

Honestly, we don’t know. I’m confident my theory accounts for *some* of these new hospitalizations, but I’m also confident it doesn’t account for all of them. And I only saw the details for this one hospital… maybe it’s different in other hospitals.

I want my theory to be right. I want it to be *the* single, simple, clear explanation for the phenomena we are seeing. It is very comforting to feel like I can contain the totality of what is happening with a simple explanatory thesis.

I can’t. You can’t. No one can do this. There are too many variables, there are too many regions, there is too much going on for us to be able to explain it with one simple reductive story-line.

Anyone who says they can explain what is going on with one single insight or in a single tweet thread or in a 2 minute “here’s one weird metric that explains everything”, that person is wrong. That kind of thinking should send a siren off in your head that the person who is saying this has a shallow grounding in the very complex world we live in. That they could actually think it is possible to figure out The Answer to all the data patterns that we see indicates to us that they haven’t thought very hard about all the metrics and how they interact with each other.

PoliMath @politicalmath

Everyone is telling me that deaths are a lagging indicator and while that *may be true*, I'm VERY wary of this as an explanation b/c it is not based on evidence but based on the fact that it's a very tidy way to explain what we are seeing. That raises alarms in my brain.

I go through all the data that is available from a half-dozen sources. I ask tons of questions from subject matter experts and I’m learning things everyday. I feel like I’m absolutely fire-hosing information in these newsletters and I’m barely covering half of really important context. And if I’m flattering myself, I might say that I know maybe 10%-20% of the context of this data.

There is a lot to know. If any given theory is obviously true, the data won’t be shy about it. You won’t have to tease it out or play guessing games or theorize about hidden causes or manipulated data. The numbers will burst right through our talking points like the Kool Aid man and scream it into our faces.

Don’t worry. When something big happens, you’ll know about it. It’s the small to medium sized stuff that we need to keep an eye on.

Why Deaths Are My Gold Standard Metric

Since the beginning of this crisis, I’ve been struggling to figure out what metric are the most stable across states. What are the best, most reliable, most stable ways to gauge what is happening?

I landed quite early on deaths as my “final word” metric. There are some reasons to be skeptical that the death count is accurate (both under-counting and over-counting) but also keep in mind that *any* metric is going to have inaccuracies. I decided deaths was the clearest one.

But why? Let me start with my reasoning and then I’ll go through the arguments against it.

I don’t want to be flip discussing the value of human life (oh God, what am I getting myself into), but deaths are a very big deal. Not just in an “infinite value of human life” sort of way, but a human death has enormous legal and financial consequences.

When this crisis began, many state health departments were simply unprepared for things like reporting negative tests. This makes sense. Why would you report the negative numbers for an HIV test? That’s ridiculous. You just need to report the positives, that is good enough. Major parts of our public health infrastructure was caught off guard by the kinds of data demands that COVID thrust us into.

But not deaths.

Death reporting is incredibly important. It is incredibly robust. We have enormous existing infrastructure to determine cause of death, time of death, to report deaths and their causes and symptoms to the CDC. One of the big “scandals” in Florida was when state officials asked county coroners to please route their death reporting through the state health department for review instead of releasing them directly to the public.

Death reporting matters. It matters for financial reasons, for public health reasons, for legal reasons. Someone who dies in the final moments of December 31st is going to be in a different CDC category from someone who dies in the first moments of January 1st. And those two people will potentially be subject to different tax laws regarding their estate.

For these reasons (and many others for which I cannot hope to fully account absent a short book) we have a pretty good medical pattern and legal structure for reporting deaths. The bigger question for this topic is around “was this a COVID death or not?”

To this end, there are a couple caveats.

Some people will say that we have over-counted COVID deaths. They have a point. In the deepest moments of this crisis in places that were most devastated by this virus, nearly everyone had it. 50% of people who were tested were testing positive. Hospitals were being crushed. Doctors were losing multiple patients every hour. So if a patient came in with a stroke and tested COVID positive, the role of a responsible doctor is not to say “well… let’s sit down and think really hard about this case. What was his history? Was this stroke COVID related? Let’s get a second opinion and consider this carefully.” The responsible doctor says “He tested positive for COVID? OK, code him in as COVID, send him downstairs, and help me with this lady over here, I think we can still help her.”

In such a scenario, mistakes will be made, but to assume those mistakes are some part of a conspiracy is to desperately misunderstand what was happening on the ground.

On the other side, there are the deaths that were probably COVID but never counted as such. These include the infected who, seeing the disaster that was on their community, decided that a peaceful death at home surrounded by their family was preferable to an isolated death in a cold hospital while hooked up to a half dozen machines. These individuals never tested positive for COVID and therefore never counted as COVID deaths.

Let me jump out ahead of things and say that the latter (the under count scenario) vastly outnumbers the former (the over count scenario). For reasons that involve the phrase “excess death” and that I swear I’m going to get to eventually, it is much more likely that we have under counted COVID deaths than over counted them. We have probably done both, but we’ve under counted more.

Even so. When we as a society, as a legal and moral culture, start looking at deaths, we tend to get very serious about things. County health departments want to know what caused the death. So do state health departments. And the CDC. And the doctors. And the family of the deceased. And the life insurance adjuster. There are a lot of interested parties with financial and legal skin in the game when it comes to death.

The result is that death statistics are fairly stable and reliable. A lot of people care (for reasons outside the infinite value of a human life) about the details of death. We’ve spent decades building the infrastructure to accurately report on deaths. This infrastructure is worth trusting in part because it bears a certain legal and financial burden that we rely on even when we are not in the midst of a pandemic.

COVID and the Protests

There is a lot of talk about protests (I’ve mentioned it just last week) and to what extent they will exacerbate COVID transmission. Chris von Csefalvey is an extraordinarily talented data scientist and epidemiologist and has been, since the beginning of this crisis, my primary resource for things ranging from the inner workings of PCR tests to viral origin to pandemic patterns. If you’re on twitter, you need to follow him.

Chris had an absolutely vital thread on the protests, COVID spread, and what we should do moving forward. It’s so important, I’m simply going to reproduce it here in full.

By now, we ought to be seeing some evidence of increasing #COVID19 cases from the mass protests. That we are not (and not even an increase in testing demand) raises very serious doubts about the #coronavirus lockdowns and other NPIs (many of which are still ongoing).

One explanatory hypothesis is weather, i.e. that outdoor protests in warm(ish) weather might have decreased the infectious potential of SARS-CoV-2. This is highly unlikely – even if there IS such an effect, it almost definitely doesn't outweigh close proximity.

Another explanatory hypothesis is that most protesters were relatively young, and therefore unlikely to fall ill but may asymptomatically transmit. I've got this shelved under 'we'll see', but I'm not 100% convinced this will indeed happen.

The third explanatory hypothesis is that we have vastly overestimated certain risks from COVID-19, limiting fundamental civil liberties of law-abiding citizens and getting it wrong. It finally took unsanctioned mass protests to prove this point.

At this stage, I find (3) the most likely explanatory hypothesis. As I said many times, a novel pathogen is notoriously hard to predict and it's better to be safe than sorry – so even in retrospect, I am not sure I would have advised significantly different courses of action. However, the current situation has unveiled an utter lack of a solid public understanding of public health policy. By necessity, public health has the power to limit fundamental civil liberties in cases of emergency – this is regardless of the merits of the subject.

What is extremely dangerous is when it appears that opinions, often right and sincerely held (e.g. the opinion that racism & police violence is a significant problem, an opinion I happen to share), influence the application of limits to those civil liberties. Or, to put it more bluntly: if public health rationales oppose mass gatherings, then they must do so across the board – whether it's BLM, anti-lockdowners or the Monster Raving Loony Party's annual open air congress. A righteous cause doesn't confer anti-viral immunity.

In this sense, America embraced public health as a common cause just until it began to disagree with something many happened to agree with *more*. The disparities this creates are incredible, and it is difficult to see restrictions as unbiased public health measures now.

From a purely scientific perspective, there's no difference between lockdown protesters and anti-racism protesters. Given that exposure risk increases exponentially w/number of people, the opprobrium that the anti-lockdown protests have engendered seems excessive in retrospect. What we have right now is the worst of all worlds – a justified notion that public health measures are, to paraphrase Anacharsis, like cobwebs – strong enough to catch the weak but too weak to catch the strong.

This is a dangerous trend. America has accepted public health imperatives not as the profound truths they are but as a cargo cult, as a way to signal and perhaps generate some sorely missing social cohesion, but with no depth and no real understanding of what's going on.

There are no winners in this. The losers are all of us.

This was the last quarantine America has ever had, or will ever have, in our lifetime. The next pandemic – we are overdue for two more in the next ten years – will burn unfettered through the population when it comes.

An abundance of caution is often a good approach, but it requires honest communication of the inherent uncertainties followed by consistent execution. Public health did okay-ish on the first – but its political executors completely failed on the second point. There will be, I'm afraid, a very, very steep bill to pay for that in years (and pandemics) to come.

I like Chris. More importantly, I trust him. He has a sense on his field and the topic of pandemics that I can’t hope to have unless I quit my job to pursue a life of public service that no one will ever appreciate. Chris speaks honestly, gives a view into his field, and is good people. I fear his warning will be prescient, and that the mistakes we make in this year will resonate in decades to come.

Disney Shorts: The Blind Mousketeers

This is one of the purest early forms of physical comedy. These three mice would have at that time been something of a tribute to the Three Stooges. We have the three blind mice bumbling their way through the traps of Peg Leg Pete (sorry! Peg Leg Pete is actually “Captain Cat” here). While the Three Stooges staged their physical comedy off each other, the Mouseketeers make Captain Cat the butt of all their physical jokes.

This is very much in line with the Disney ethos at the time. Early Disney was more comfortable with poking fun at their protagonists (Mickey was a cad for his first few years in the black-and-white era). There is a weak attempt to manufacture some drama as we risk the mice to the cruelty of Captian Cat, but most of the comedy is in how the blind mice evade the elaborate trappings of the authority figure and manage, through luck and wits, to overcome his tyranny.