A White House official says the Coronavirus Task Force “will be phased down around Memorial Day” (CNN Politics). WTF?
15 Children are hospitalized with mysterious and extremely rare illness linked to COVID-19 in New York, some European countries (New York Times)
Asia passes 1 quarter-million mark in cases, but has just 7% of the global burden, compared with 40% in Europe and 34% in North America (Reuters)
First of 4 candidate vaccines from Pfizer begin human testing (Washington Post)
Alaska received $3.4 million in federal CARES ACT aid per COVID-19 patient, New York got only $24,000 per case (Forbes)
Is there a more contagious second “strain” of coronavirus? Not necessarily I feel the strain of all this talk about strains. Let’s call it strain strain (or strain2 for short). You may have heard about a story published in the Los Angeles Times today with the headline “Scientists say a now-dominant strain of the coronavirus appears to be more contagious than the original” (article behind a paywall). Searches for “mutant coronavirus” shot up 5,000% according to Google Trends. This story was followed by fierce outcry among scientists on twitter. It’s based on a report posted on BioRxiv by scientists at Los Alamos National Laboratory. The study has not been peer-reviewed, so let’s approach with caution. Informal review is now happening on twitter. The article is by computational biologists who used over 6,000 genetic sequences of the SARS-CoV-2 virus from around the world, looking at subtle variations in the code. They focus on one mutation called D614G which they say emerged after the virus first broke out in Wuhan and has since become more common in hard-hit places like Italy and New York. They say this is enough evidence to warrant an “early warning” that this constitutes a new “strain” of the virus and one that must be more contagious because it seems to have taken over and spread faster than the original strain. Not so fast, say the experts. After a deep dive into the tweet storm, and communicating with several experts, I am convinced that the LA Times story over-states what’s actually in the Los Alamos report, and rushes to judgement in a number of important ways. I have said before that experts in the field have been careful to point out that mutations occur constantly in a new virus. Whether those mutations change the way the virus works (in terms of its transmissibility or lethality) is a different matter. It’s too soon to tell if this mutation has made the virus more contagious, or whether it just got lucky and happened to spread faster due to what experts call “founder” effects or random genetic drift. Normally, when a mutation is found like this one, using studies of genetic data alone, scientists have to do followup lab studies to confirm that the mutation in question is actually doing something or is just random window dressing. That usually means animal research under careful control to zero in on the mutation’s function. Nothing like that has been done yet. Therefore, it is possible that this, and all other mutations that have occurred, is just luck-of-the draw for the virus and not a signal of a new and worse “strain”. Of course, the possibility of a second more contagious strain also cannot be ruled out. It’s just that we don’t have enough to conclude, as the title would suggest, that “scientists say” there is a second strain. Anyone interested in the details can look at the twitter posts from Dr. Bill Hanage from Harvard’s T.H. Chan School of Public Health (@BillHanage). His bottom line: “…right now there are better ways of fighting the pandemic than worrying about different strains”.
U.S. deaths spike, set new daily high. Weekend reporting lags skew the picture On Tuesday, a record high 2,627 Americans died of lab-confirmed COVID-19, a rise of 7%, calling into question the trend toward declining deaths. Seventeen states matched or exceeded their previous high daily death count: Arizona (21), Colorado (37), Hawaii (2), Wyoming (4), Indiana (61), Michigan (232), Minnesota (17), Nebraska (5), Oklahoma (21), Alabama (31), Kentucky (17), Mississippi (14), North Carolina (34), Delaware (10), Maryland (68), New Jersey (376), and Pennsylvania (360). Making sense of the up-and-down of deaths continues to be especially difficult. I am now convinced that we are seeing a pattern in our reporting over the past few weeks that contributes to the confusion. I believe we are seeing a consistent lag in the reporting of deaths over weekends. Below I show you my daily death counts, but there is something new. I darkened the color of Sunday and Monday reports (darker pink). The pattern is striking. The ping-ponging in deaths may be a function of a lower ascertainment rate over the weekend. It’s not apparent on Saturday because those are deaths occurring Thursday and Fridays. Over the last month, reports issued on Sunday and Monday are dramatically lower than the rest of the week. Tuesday reports have been correspondingly spiking as administrative paper work catches up to the weekend lag. To be sure, some of the inconsistency results from states injecting new deaths as big batches of test results coming in, or reporting procedures change. If I am right about this pattern, we keep getting seduced by an apparent drop in new deaths on Sunday-Monday, only to be disappointed by a mid-week resumption of increasing deaths. What does it mean? It is challenging to put this unwelcome news into context. As I have said recently in this space, it is important to look not just at daily numbers but at larger trends. Let’s see if the longer term trend is flat growth in deaths or continued rise. In the meantime, we need to pay attention to those states now with elevated epidemic intensity (see lower graph). Pennsylvania, for example, has seen 87% growth in deaths over 3 days.
Good news: with all this uncertainty, there is something really important that we do know In what I consider a deeply unfortunate and ill-timed opinion in the Los Angeles Times today, Mariel Garza opines that “everything you know about the coronavirus is wrong”. Well, she is wrong. It is true that every day, we are confronted with things we don’t know. Can I get reinfected? What’s the real case fatality rate? If I have antibodies, am I immune? When is this going to be over? Lots of uncertainty to be sure. The good news is that we do know something really important. I have hesitated to bring this up, because it requires some math, but I think it’s time to explain something very powerful that we do know. The topic is this: when and how will we know when we have hit peak in the US? The term peak is thrown around all the time, but what does it mean? Before I get there, consider this: we do know what the epidemic curve for this outbreak looks like. First, let me show you what that curve does not look like. Below are 3 bone-head-simple graphs of what the epi-curve could look like in theory. Figure A shows cumulative growth in new cases for the linear model. In this graph, cases grow along a straight line until there is nobody left to infect and the epidemic stops. Unless you have been under a rock, you already know that this is not the reality we are living in. Every country hit with the epidemic has seen non-linear growth of cases. It’s actually been exponential in every country at least for a time. Figure B shows what the epi-curve would look like under an exponential model of cumulative case growth. When we look at countries that were first to be impacted, like China and now Italy, we know that this isn’t our reality either. Epidemiology 101 tells us that at some point, there are not enough susceptible people left to fuel the outbreak and the rise of new cases slows and eventually stops. In nature, exponential growth of anything can’t continue indefinitely. Based on the way epidemics behave everywhere, we know that the COVID-19 outbreak is following a logistic model (Figure C). That implies three distinct phases: initial exponential growth, followed by linear growth, and then a final phase of declining growth. We have the math to describe this curve pretty accurately. The problem is that the data we have at any one point is not good at telling us where on the curve we actually are. Due to something called the “base-rate” problem, predictions based on this curve are very dangerous very early in the epidemic when the prevalence is low because the disease is still rare. That’s one reason models are so wildly variable. We know that the key thing we should be looking for is the inflection point (see Figure C) in the middle of the logistic curve where exponential growth transitions to linear growth. If we can identify that point, we know that something really important has changed. We know that we are about half way to peak in cumulative cases. We know the rate of change in new cases has peaked and the outbreak is slowing. Back to the key question: what is peak and when do we know we are there? First, peak is not the top of the cumulative case curves. That won’t happen until the epidemic is “over” (which is not likely for another 6-18 months). The peak that matters, and that we know will come, is the point at which growth in new cases shifts from exponential to linear, at which point, new cases will shift from increasing to decreasing, and the rate at which people are recovering will exceed the rate of new infections. So what should we look for? In order to see the inflection point, we have to look at growth-rate in cases in the logarithmic scale to see where new case change flattens. Not new cases, but new case growth. More precisely, the best thing to look at is a graph that shows log of new cases over a week plotted, against all previous cases, both expressed on the log scale. This log-log plot doesn’t use time on the X-axis for complicated reasons. The inflection point shows up very clearly as a sharp drop in the growth trajectory. The bottom graph comes from a terrific website created by Aatish Bhatia that computes this exact graph for many nations. The green line (my annotations) shows just how similar all nations are during the period of exponential growth. The green circles show the tell-tale sign of the inflection point, clearly visible in Hong Kong, China, South Korea, Australia and Taiwan. The bottom line: We know something very powerful about this epidemic; we know what the epidemic curve’s shape will be, because we know what “model” is underlying the epidemic. We need to look at the data the right way to know when the inflection point has happened. Peak growth in new cases means the rate of new infections has shifted. Reopening the country before this is clear will lead to an unnecessary re-ignition of infections. The figure shows the US is now wavering. The period of exponential growth is slowing and possibly nearing an end. However, it suggests peak in rate of growth in new cases has not yet occurred; there is no clear drop. The US has not yet peaked. We do, however, know what to look for moving forward.
Screen grab taken on April 21 of an interactive graph created by Aatish Bhatia in collaboration with Henry Reich at https://aatishb.com/covidtrends/. Figure based on data from Johns Hopkins CSSE and the New York Times. Green stuff are my annotations of this slide.
What a difference a month makes, the pandemic moves east Every once in a while, I like to wind the clock back and see where we were a month ago. Yesterday I glanced back at March 21. The two figures below show just how ‘long ago’ that was. The pie graph on the left shows deaths by region yesterday. The Northeast region has been the clear epicenter of the outbreak last month with 64% of all US deaths. One month ago, the US had recorded less than 300 total deaths! Yesterday, we saw nearly 3,000 deaths in 24 hours. A month ago, the largest fraction of deaths was in Western states (43%). I doubt anyone would have predicted on March 21 that the epidemic would shift so dramatically toward the northeast. Not many believed then that we would lose 40,000 Americans in just 1 month. But, that is what happened. That itself doesn’t tell us where we will be on May 21, but it does remind us that tomorrow’s hot spot may not be the place we are focused on today.
Thomas A. Glass, Ph.D. 