COVID: Stop Pretending we Know Things

Every day there’s another viral article or study. Sometimes it’s a reason why we’re all screwed and this will be a worse death toll that any world war. Or, more frequently over the last few days, the message is “don’t worry!”: a “growth hacker” or tech CEO / VC looked at a bunch of charts and says there’s nothing to panic about.

Unfortunately, it’s those articles and studies that are most aggressive in their claims that get a ton of distribution as they play into the confirmation bias of either the doves or the hawks. Instead, we should be reading balanced things about what we know and don’t know.

For example, today’s viral medium darling is “Evidence over Hysteria”, a horrific mishmash of charts and data points (good data actually) that ignores anything even in its own charts that’s concerning and makes a bunch of claims that are self contradictory. The main issue here is that for every data point presented, not even the slightest attempt is made to look at “oof, how might that be misleading?” or “does this really say what I’m claiming?”

This is just bad analysis. We need to think deeper and be honest with what we don’t know.

Do I know what will happen? No! But I can tell when plainly contradictory facts are put in front of me. And I can be honest about how much is still unknown.

Here’s my summary of that article. And the sad part is that most things the author doesn’t mention are evident from a chart just a couple scrolls down that same article (or is on “Our World in Data”, a great source, just down the page from the chart the author used).

  • Don’t panic because the JHU map has “total” cases as the headline number, but, 1/3 have already recovered so no worries! We get 33% off.
  • Don’t worry that US is tracking 11 days behind Italy on total cases because US population is 5x bigger, so per-capita our total cases are 5x smaller. No mention or discussion that this may mean we peak 5x higher. Or that means we may have 5 or more Italys — one in NYC, one in NOLA, one in Seattle, and let’s see where else.
  • Don’t worry: US has only 60 cases / 1M population, whereas European countries have 200–700. No mention that as testing ramps, found cases doubled in 2 days. Are we really 10x better off or just 6 days behind?
  • Don’t worry: Daily growth rates declined over time in every country. Generally good news. But, no mention that for now, many countries are flat-ish rates at 10–20% daily growth. That’s doubling every 3–6 days. USA is at 25% and not slowing much yet, likely because testing hasn’t yet caught up. The only countries that truly wound up at a slow rate are SK (highest testing in world) and China (locked everyone in buildings for a month). Literally the next graphic in the article (!!) contradicts the point. It shows that world growth rates have either accelerated or stayed flat day-to-day for the last 30 days in a row, including accelerating the last 4 days. It was Feb 21 since we last saw a drop. This is just irresponsible.
  • Don’t worry: all epidemics follow bell curves, whether due to environmental or our efforts. Luckily, this is true. However, the trillion dollar and the million of human lives question is: is this gonna be one that just fizzles out on its own at a very low population penetration rate? Or is it going to be like the flu, which infects 40–60M people in the US per year before reatreating from seasonality and herd immunity? There’s no mention or attempt to wrestle with this question. Instead, we are presented the China and South Korea bell curves, who seem to have beat it back with super aggressive measures (mix of lockdowns, masks, testing, tracing). There’s an Italy graph presented from Mar 15, which is cited as evidence that Italy is also on a bell curve and things are fine now. It literally says: “don’t look at the purple area, because that data hasn’t come in yet.”

So how did this brilliant analysis fare? Well, luckily, that was from Mar 15, showing a drop in cases on Mar 12, 13, 14. So we can see how we’re doing now on Mar 21:

Oopsy daisy! Instead, Mar 14 jumped to ~3500 cases from ~2300, and by Mar 20 we’re up to 6000 cases per day, triple what the graph was showing. Let me superimpose these with up to date data. Yeah, really nailed it.

  • Don’t worry: you have a low probability of catching it, 1–5% per outside world contact, and 10% in the household. However, no mention that later in the article, it says that family transmission was the top vector in the horrific Wuhan situation.
  • Don’t worry: the flu led to 17 million medical visits last year, so this is tiny in comparison. Only 1% of cases are severe. No mention that this is a now widely debunked trope that flies in the face of the situations medical facilities are facing in Italy, in New York, in Washington State, and elsewhere. Here’s data from Italy on ICU demands from flu (this data reflects not all influenza-related deaths, but those diagnosed in ICU). It already looks bad through week 11, with 1000 people in IU and 679 deaths.
  • Here’s the worst part: let me extend that graph. In week 12 (Mar 12–18), deaths are up to 1800 (red bar 3 times taller). In week 13 (Mar 19–25), things are grim and we’re on a path to 5000+. 793 people died on Sat 3/21. The Italian army is having to send canvas covered trucks to take away all the coffins. NYC is a few days away from a similar situation. At this point, the flu comparisons are just insulting and dangerous. Flu? Are you serious?
  • (Note: Even if we add in the most aggressive flu models for all downstream deaths — not in ICU but looking at total excess mortality, we get ~22k deaths per year or ~1000 deaths per week. We just blew past that 5x. But who knows what the same methodology applied to COVID-19 is — excess mortality from downstream effects will be huge for this epidemic.)

The rest of the article is just an equally bad discussion of fatality rates. There are a lot of unanswered questions which are presented with comical confidence. The implications of contradictory data (e.g. that SK has a TON of young people cases, as do Germany and Italy, whereas China does not, are ignored.

We need real answers, not this BS. We need to know how many asymptomatic cases there are, because then we’ll know whether the Italy and NYC curves will continue until symptoms hit ~30% of the population whether it will top out 10x smaller or even 100x smaller than that. We need to know whether children can actually transmit or not (studies are mixed). We need to understand whether we’re over-applying the Diamond Princess data (aging population, where 20% got it and 1% died) — can we project that out to broader populations? When looking at promising treatments like hydroxycholoroquine, did Korea and China already use them at scale and if so, are all the benefits already baked in to the Korean fatality rate? Why is Germany so hugely different in both age mix and outcomes to Italy and Spain? What would the South Korean trajectory be without extensive testing and masks? There is so much we don’t know.

But as we look at this metrics, we need to not willfully close our eyes to harsh realities on the ground. Italy is going to cross 1000 deaths a day in a day or two, and we likely will too. It doesn’t take many days at that rate to match and surpass the annual~20–50k deaths from the flu in the US. For an article about exponential growth and mocking “false metrics”, there sure is a huge disconnect to real life here.

It is totally right to question the predictions and panic. It is totally right to ask about the economic devastation of lockdowns and how this will also kill and hurt the health of so many people, especially in this country, who have no cushion and can’t absorb a period of unemployment. It is totally right to ask if school closings are a benefit or a cost. But it needs to be done with an honest look at the things we DON’T know, and contradictions must be attacked and exposed so we get to the true answers.

Eng & Product VP @ FB. Loves: awesome managers, people who grow, tech, food, design, sports. Follow me as @mrabkin.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store