Quantifying COVID (w/ Balaji Srinivasan)
ASH BENNINGTON: Right, and, to pick up on that, where do you think we are now? We're recording here at the end of April. Where do you think we are in terms of the state of play with this virus right now in the United States and globally? BALAJI SRINIVASAN: Well, so there's a really good site called covid-- so there's two sites that I look at. The first is covidtracking.com and all the sites that use that data. And covidtracking.com is looking at tests, cases, deaths, et cetera. And it's a really good project by Alexis Madrigal and company. And, basically, they were just finding that the CDC wasn't able to track tests. So they went and it themselves. Massive props to them. That site is updated very frequently, very labor intensive effort. And it's pulled by a bunch of other folks. And, if you go to a website called covidcharts.com, basically-- you have to wait until the end of the day. But the new cases per day and new deaths-- all right, so, looking at April 27, it does look like it's down, at least as of this morning or whatever, but you have to wait a few days to see that. In general, it doesn't look like we have massively decelerated yet. It doesn't seem like the deaths or cases have come down. But people are like, oh, it happened in New York City, and so on. Well, actually, it happened in New York City, and it came down in New York City. But it's in 49 other places, 49 other states. So I think we're just at the beginning of this thing, unfortunately. And I think that people who think, oh, the pandemic is over, oh, it only happened in New York City, it's going to come to their town, potentially. And ASH BENNINGTON: How do you think about the trajectory of the right now in the United States, just from a disease-impact standpoint? BALAJI SRINIVASAN: Yeah, so one thing that-- so a few things I think are clear-- and let me at least give you my world view on it because there's some very big variation in people's thinking on this. First is I'm pretty skeptical of the serology work that came out of Stanford and USC and others. I wrote a long Medium post critiquing it, and so did a bunch of other people because I think the false-positive rates of those assets are very high. Why is that important? There's basically been this thesis that, oh, a bunch of people have already gotten it. Because they've already gotten it, it's already spread a lot. That means we're close to the end. We're already at 20% or whatever, 30%. Just a little bit more, and we're almost done with this. And the reason I think that's unlikely is it would mean that this virus is way more contagious than other pandemics. Like H1N1, or the Spanish flu, took more than a year to get around the world. H1N1 was in the age of modern jet travel. i be spreading really fascinating. I think that's unlikely. And I also think that those kinds of calculations don't really contend with the second major metric that I think is important, which is this guy-- his name is John Burn something from the-- John Burn-Murdoch from the Financial Times. J Burn-Murdoch, B-U-R-N, not B-R-N-S. So that guy has a very useful set of graphs that is something I'd been talking about since March or April. And they took that next level. And they've created graphs of the spike in death rate over the historical average, so the spike in all-cause mortality across a dozen countries. And the thing is that I start from the premise that this is the best test. Unfortunately, COVID-19 has now got into the scale that it is a very significant cause of death in more than a dozen around the world. Actually, way more than a dozen, by at least a dozen for which we have good data. And because it is at that scale, that means that you can compare it to past years. And you can see how many excess deaths there are. And the reason that's important is death is-- well, from a diagnostic standpoint, I don't mean this in a callous way. But death has high test-retest reliability. It can be diagnosed outside a lab. You don't really have false positives, and so on and so forth. So lots of the issues that bedevil other kinds of diagnostic tests are not there with death. And it's also something where you have probably the best records that you might have. If you talk about, oh, does somebody have a pneumonia? The diagnosis of pneumonia would probably differ from country to country in terms of like what a doctor would call pneumonia. But death doesn't differ from country to country. So it's the most consistent across space and time. It's easier to diagnose. So it's really the best stat, unfortunately. And those stats are showing just a pileup of bodies. There's a huge surge of mortality in not just Italy, but Spain and a bunch of other countries, including New York City. And J. Burn-Murdoch has that. So that's the second set of charts that I look, the all-cause mortality charts. ASH BENNINGTON: This was something that came into sort of the broader consciousness, I think, yesterday, where, effectively, there were some articles saying that all-cause mortality seemed to suggest that the death rate from COVID was twice as high as was being reported from other diagnostic criteria, which is obviously an enormous gap. BALAJI SRINIVASAN: Well, yeah, so I was just saying, though, that not-- definitely not to thumb my own chest, but because-- if you look at my Twitter, we've been talking about this since March and pulling up analyses that were kind of-- So there's a guy, Dellanna Luca, was doing these analyses in Italy. There's another guy, David Bessis. So kind of tech Twitter or whatever has been kind of digging into this for a while now. And so we had already kind of had the numbers to show that spike and all-cause mortality like a month ago. But great credit to John Burn-Murdoch because he's taking it next level and actually gone have pulled it for even more countries and gotten the government stats in a bunch of different places to kind of solidify the analysis. So let's call that basically just anchoring the severity of disease. Now, who is it severe for? The short answer is it's very severe for, A, old people, meaning 65 plus in particular-- but, as you get older, just risk increases-- and, B, people with pre-existing conditions, OK. But the second bit should not be that reassuring, especially to Americans because they have a pretty high rate of obesity and heart disease and so on and so forth. It's an amazing number of unfortunate Americans who have one or more of those comorbidities. So people who are like, oh, you know, like that's not me. And they're 53 and overweight. Well, they're kind of in a relatively higher-risk category. And you can look at the stats on it. It differs from study to study. But the CDC had some numbers on this. And they did a whole demographic workup. The other thing, though, it's not that people are immune when they're younger. Something like, if I recall correctly, 25% of hospitalizations were for those under 50. So 75% were over 50. But 25% were under 50. And that's not 0%. Some of the stats other states have seen is something like 13% hospitalization rate for people 18 to 44. And I know people who've been very sick from it, just kind of qualitative evidence to buttress that. By the way, on the topic of anecdote, by the way, people say, well, the plural of anecdote is not data. And I would agree except that qualitative feedback often allows you to formulate a quantitative hypothesis. And that's why it's actually super important in tech startups, just to digress for a second, to talk to customers as well as look at your dashboards because you talk to a customer. And they say, oh, I had trouble logging in. Then you put up a new dashboard, see how many people had trouble logging in. You see 1,000 people have trouble logging in. So the, quote, anecdote led you to collect data that-- So we should never discount the value of a patient study. That's actually really, really important because it lets us, maybe, look at things from an angle we weren't considering and then collect data to see if that observation is real. So point basically being, A, I think it's a quite severe. B, it's severe in the elderly and people with pre-existing conditions. C, that doesn't mean the younger people have clean sailing of it. And, D, something else I'm concerned about is folks who were so-called asymptomatic. Many of them appear to be not asymptomatic but pre-symptomatic, meaning that you do follow up on them, and suddenly you're seeing CT scan of their lungs is actually showing evidence of viral damage. And now, they start coughing, or they have other kinds of symptoms. And so, for at least some people, the virus is like a slow burn, which is bad because it's kind of hiding. And so there's other things where the sheer number of different kinds of symptoms this thing is causing in folks, we're still kind of getting the table. People are reporting strokes. They're reporting weird lesions in the toes. Whether this is due to simply just the scale of it, causing lots of rare cases, whether it is due to it actually mutating and there being different strains doing different things that some people have speculated, we don't really know yet. But I would just treat this thing as being fairly severe and not something you want to mess with. And so I think the strategy of like in Singapore or New Zealand is probably right. OK, I know that's a long answer. But go ahead. ASH BENNINGTON: I'm so glad you brought that up. The two sorts of things that I was really trying to reconcile from yesterday was the report of the double all-cause mortality based on some of the analysis that we were just discussing. And the second point was there was a study coming out of the state of Ohio, I believe, about testing prisoners, showing that 96% of the prisoners they tested were, in fact, asymptomatic and trying to reconcile those two propositions, on the one hand, something that would suggest the disease is much worse on the surface than it appears, and then the second, suggesting the opposite. It is a very complex problem to get our heads around. BALAJI SRINIVASAN: Yeah, and I think-- my speculation is that those asymptomatic prisoners, many of them will develop symptoms. I tweeted this a little while ago. But, in a follow-up study of the different kinds of follow-up studies of the asymptomatic, like 27%, 54% in different studies develop symptoms at some point. Yeah, I think that's what's going to happen with those prisoners, that they're not going to remain asymptomatic indefinitely. Not all, but many of them.