At the time I wrote that there were 14 known positive cases diagnosed of COVID-19 in the U.S. But as I said, that’s not how the epidemiology of an outbreak works mechanistically: It was instead that we hadn’t correctly identified all of the other cases around. So as of today, there are 33,404 cases positively identified in the U.S. And that won’t be all, either. Our limit has been on the number of test kits available and in-use, which was literally about a dozen per day a few weeks ago—and is about 40,000 per day now. Given the number of reputable and reliable diagnostic labs in the U.S., we should be able to support a testing throughput of between 100,000 and 150,000 tests per day.
Welcome to the Recession
On Friday, March 20, Goldman Sachs projected that GDP would fall at a 24 percent rate in the second quarter. Eight days earlier, on March 12, the stock market had its worst crash since 1987. These are trying times, and in a global market, epidemiology doesn’t occur to the exclusion of econometrics.
At this point, a recession is all-but guaranteed. Greg Daco, chief U.S. economist at Oxford Economics, stated we may be positioned for the biggest quarterly market decline on record, noting, “This is not just a blip, we’ve never experienced something like this.”
Literally innumerable small businesses (and even not-so-small businesses) have laid off or furloughed millions of workers and many have closed out of fear of the unknown. In their cases, it’s the unknown of when government mandates to socially-distance will abate.
It doesn’t help that the Surgeon General of the United States, Jerome Adams, just commented live in an interview on the TODAY show that “…this week, it’s going to get bad.” He meant it to address the vast majority of people who didn’t take social distancing measures seriously over the past few weeks. But it was the exact opposite of helpful discourse for a nation (and world) already on edge with the news media’s scaremongering coming in a continuous stream from all angles.
The Harmful Contribution of Social Media
Facebook, Twitter, and Instagram have been alight with activity, and are all bastions of anti- and pseudoscientific rhetoric, within which the respective company heads are currently combatting potentially harmful messaging. Unfortunately, finding all of the incorrect content amidst the surge is impossible, and Facebook, for one, is resorting to using AI to do the locating and identification, because the vast majority of their workers are themselves trapped at home. Our hyper-connected world makes pandemic response paradoxically somewhat worse at the moment because of the rapidity of the spread of MISinformation.
Human Behavior Under Panic
Typically, when faced with a crisis, people in groups exhibit what’s called a Therapeutic Community, where they are very altruistic at the early stages of a disaster. Contrast this with one of the most unexpected incidents over the past month: The nationwide hoarding and national shortage of paper products—toilet paper, followed by paper towels, and facial tissues.
Here’s what’s going on. There are 3 major paper products producers in the U.S., who themselves rely on binder glues produced in China. As the national producers of paper goods stock-out of product, they look to ramp-up production of the next batches, which is now compromised due to a manufacturing shortage of raw materials for the process. What’s interesting about this situation is that there seemed to be 3 distinct strata of people engaged in this hoarding behavior. The first group reported that they were fearful of apocalyptic outcomes, and wanted a private stockpile of supplies. The second group noted that they weren’t as concerned about SARS-CoV-2 per se, but that they wanted some extra supply to not need to go out in public as often. And the third group observed that they weren’t afraid of the current situation, but the panic-buying of the other two groups drove them to participate in the buy-outs because they didn’t want to be left with no supplies remaining. The precise recipe for scarcity of essential and nonessential goods during a public panic. As of the time of writing this article, there is also for example, a severe shortage of infant formula in Maryland (as well as other cities) for these exact reasons I’ve laid out above.
What we actually have is a Public Panic Pandemic, where the information fed to the population and their response to it, has created a pernicious cycle that’s leading to worse outcomes than the coronavirus itself.
Why Does This Behavior Exist?
We’ve been predicting an outbreak like this, and frankly worse, for decades. But nothing appreciable has really been done to set up systems to have prevented COVID-19 outbreaks. A lot of this is due to the sociological effect of the Status Quo Bias, wherein the public’s ritualization of behavior is a very difficult thing to break. Clearly this is an evolutionarily-adaptive feature of humans. When we follow heuristic scripts day in and day out, it not only requires less cognitive energy to maintain, but also because by its very nature of persisting, successful routines tend to be successful. We are rewarded behaviorally by having routines that tend, on average, to produce good outcomes. And so, the cognitive load to shift these, especially dramatically as is the case with social distancing, isolation, and shutdown orders, is extraordinarily difficult for many people. We see Status Quo Bias very often in behavioral economics, where the tendency is very strong to prefer the current situation, and any deviation or departure from that state is often perceived as a loss—even if objectively it isn’t.
Similarly, what sociologists call a Normalcy Bias causes people to believe that things will function in the future the way they have experienced them in the past, and therefore to underestimate both the likelihood of a disaster and the consequences of the possible effects. This is the backbone for the public being woefully unprepared for ‘Black Swan’ events, popularized in Nassim Nicholas Taleb’s titular book.
Let’s never, amidst this or any other social panic, forget the Scientific Method and the value of empirical evidence and hypothesis testing. A quick refresher for you:
The Scientific Method
1. Define a question
2. Make observations
3. Form a falsifiable explanatory hypothesis
4. Conduct an experiment to test the hypothesis
5. Analyze and interpret the data
6. Reject or Fail to reject hypothesis
7. Publish results
8. Reproduce (others)
You Want the Public to Take What?!
President Donald Trump commented publicly on March 21 that the combination of hydroxychloroquine and azithromycin—both in caps as is customary for his tweets—would be an effective treatment for COVID-19. “It may work, it may not work. I feel good about it. That’s all it is. Just a feeling.” Please see the Scientific Method above again. We don’t do science by ‘feeling.’ Untested ideas are not necessarily a good thing, and are potentially harmful.
Further statements by the President were that evidence for chloroquine’s effectiveness is “very strong,” (which it’s not) and also, “Why should we be testing it in a test tube for a year and a half when we have thousands of people that are very sick, and we can use it on those people and maybe make them better?”
Let’s break this nonsense down. First of all, there currently exist no robust data to suggest that chloroquine would have an antiviral effect against SARS-CoV-2, nor does its mechanism of action suggest to me that it even should be effective. And azithromycin, which as an antibiotic, would have no activity against a respiratory virus. The secondary effect of this reckless recommendation, which is actually costing lives and quality of life now, is that the public has been rushing in quoting this dangerous hyperbole to their physicians and getting prescriptions to try this untested and likely ineffective combination. And both of these treatments can be toxic, without any known benefits in this situation. The illusion of safety is worse than the awareness of danger.
There are now many, many case reports of clinics being stocked-out of chloroquine for those patients who actually need it to modify the course of their own disease, such as lupus. ProPublica’s Charles Ornstein observed, “Healthy people are stocking up just in case they come down with the disease. That has left lupus patients, and those with rheumatoid arthritis suddenly confronting a lack of medication that safeguards them, and not only from the effects of those conditions. If they were required to take stronger drugs to suppress their immune systems, it could render them susceptible to more serious consequences should they get COVID-19.”
This idea that trying anything is “better than nothing” is entirely incorrect. FDA’s Right To Try and Expanded Access (“compassionate use”) can indeed help patients, but allowing the reckless experimentation of any treatment which comes to mind is more likely than average (actually, hundreds of times more likely than average) to lead to the attempted ‘treatment’ posing a greater risk to the average of the population than the disease for which it was designed. Former FDA Commissioner Scott Gottlieb has just recently described an approach to test existing therapies without using placebos in control groups to increase the velocity of regulatory approvals. Keep in mind that this approach is not without its own inherent risks.
The WHO is supporting a global megatrial called SOLIDARITY, which is focused on a large, diverse, observational experiment on human subjects with COVID-19 using 4 therapies: The antiviral remdesivir, chloroquine and hydroxychloroquine, a combination of two HIV drugs, lopinavir and ritonavir, and that same combination plus interferon-beta, an immune system messenger that can help disrupt viral replication.
The Important Tipping Point
Which leads me to describe the worst outcome for the COVID-19 pandemic. It appears to me at the moment of writing this, that we are inexorably past the point where the increase in morbidity and mortality on the population of existing (non-COVID-19) patients is GREATER than that likely to be experienced from SARS-CoV-2 itself. We have swung the pendulum with such force in the pandemic direction, that those cancer patients awaiting chemotherapy or surgery are being told to wait longer, rheumatoid arthritis patients cannot get their treatments, lupus patients cannot get their immunomodulators (like chloroquine), cardiac patients are getting deferred from serious interventions. As we wait breathlessly for the pandemic, this approach broadly is having a tremendous effect on patient diseases, current and future deaths attributable to being deferred or canceled because of all the holds and shutdowns, and a significant decline in Quality Adjusted Life Years (QALY).
When in Doubt (Don’t) Watch the Media Reports
The numbers that are being reported minute-by-minute and day-by-day, nonstop during this crisis, aren’t “really” the numbers. For example, CNBC just reported on March 23: “New York Coronavirus Cases Surge 38% Overnight to 20,875.” This is clearly meant to drive clicks, readership, and panic. This isn’t an indicator of the new spread of SARS-CoV-2, but is certainly an indicator that more people have been tested and confirmed positive. These are cases which became infected days or weeks ago. I think more helpful for the future of reporting pandemics would be that the adjusted—and continually-adjusting—moving average of proportion of positive cases in the population is what is primarily reported. But that requires better, high-quality testing to get a better idea of Case Fatality Rate as early as possible.
Most cases are undetected. That is still the case. The estimates range widely for how many cases haven’t been detected or reported yet, from 10x the number of current cases, to hundreds or even thousands of times the current reported cases. We know a lot about coronaviruses in general. And in this particular strain, that the R0 (“R-nought”) for SARS-CoV-2 is about 2.2. Based on 18 initial cases in Washington state and their genome sequencing, Trevor Bedford and his team of epidemiologic modelers working at the Fred Hutchinson Cancer Research Center, suggested that there were potentially 1,100 active cases (with confidence intervals spanning ~210 to 2,800 cases) ongoing in their state. This gives you a sense of how much we aren’t detecting at the moment. If we don’t know the actual population of patients with SARS-CoV-2, we can’t possibly know its actual Case Fatality Rate.
The (Mis)-Conception of Case Fatality Rates
When we talk about Case Fatality Rate (CFR) in epidemiology, the calculation is very straightforward. It is the number of deaths divided by the total number of confirmed cases:
So, using the above equation, if there are 10 deaths (numerator) in 10 confirmed cases (denominator), the CFR is 100%. That’s exactly why you may have heard that we’re trying to “increase the denominator” of tested cases. Increased testing of coronavirus makes the “denominator” grow as a critical effect.
If the number of deaths of a particular etiology stays fixed (let’s say, at 10), but the denominator (confirmed positive cases) grows rapidly to 100, 1,000, and 10,000, the Case Fatality Rate drops precipitously from our initial 100% estimate to 10%, 1%, and then .1% (for the three denominators given).
The problem is the interpretation of this number in the context of many outbreaks, including the current coronavirus pandemic. As the media reports tens of thousands of ‘new’ cases, which are really not-so-new cases, but for which we have just received confirmation that they tested positive for SARS-CoV-2, we begin to learn more about the total confirmed cases throughout the population. In this way, we are not seeing SARS-CoV-2 epidemiological vectors as they are ‘now,’ but rather we are seeing how they infected people in the short-term past. Similarly, with how we see distant starlight, it’s not how the stars are now, but how they were eons ago when they transmitted the light.
Clearly, as we approach 100% analytical testing of all people, we’ll begin to learn the true prevalence of SARS-CoV-2 in the population. And probably, too, that the CFR is much lower than initial estimates had placed it. In fact, in his most recent paper, Stanford professor John Ioannidis suggests that estimates of a CFR of 0.025% to 0.625% currently seem accurate among the groups for which we have data.4 Because using observational data from nonrandom population samples is about as epidemiologically-inaccurate as one can get, Ioannidis conceived of finding an artificially-closed population to assess.
The Diamond Princess cruise ship, he suggests, is one way to assess quarantine and case fatality rate in a more controlled way. He observed, “The case fatality rate there was 1.0%, but this was a largely elderly population, in which the death rate from Covid-19 is much higher. Projecting the Diamond Princess mortality rate onto the age structure of the U.S. population, the death rate among people infected with Covid-19 would be 0.125%. But since this estimate is based on extremely thin data—there were just seven deaths among the 700 infected passengers and crew—the real death rate could stretch from five times lower (0.025%) to five times higher (0.625%).” What we still know is that there is substantial heterogeneity in deaths across age groups, with those aged 70-80, 80-90, and 90+ having the highest risk of mortality due to COVID-19.
But even test results—positive or negative—don’t address the question as to the associated sensitivity and specificity of the coronavirus test. Each test—any medical test, actually—is mired in the reality that uncertainty exists as a result of the effect of probability in the natural world. These unknowns are, of course, the Type I error (False Positive rate) and the Type II error (False Negative rate). So, when a SARS-CoV-2 test is administered across a broad group of patients, a certain calculable proportion will return ‘positive’ results when they were, in fact, negative (the so-called False Positives) and others will return ‘negative’ results when they were, in fact, positive (the False Negatives). The corollaries of these values are the sensitivity (the ability to correctly identify the “True Positives” with the virus) and specificity (ability to correctly identify those without the virus—the “True Negatives”). Currently, Chicago-based Surgisphere, as an example, offers a diagnostic kit which claims a 93.7% sensitivity and 99.9% specificity. Many other test options appear to be less-accurate. Zhuang and others have suggested that the Positive Predictive Value (PPV) of certain coronavirus tests are as low as 19.67%.3 Improving the tests matters, because telling even 0.1% of patients tested that they don’t have SARS-CoV-2 when they actually do, is 100 infected patients out of 100,000 tested who were told they don’t have it.
Until When Must We Endure?
The way some of the models are looking, a personal cache of which I’m updating on a daily basis, is that the viral vectors won’t break now until we have a sufficient number of people in the general population who have been infected by SARS-CoV-2 and become immune, and/or an effective vaccine becomes available. With vectors which are this full (meaning, spread at the community level, and disparate cases across all 50 states), it’s likely that this virus has been circulating undetected for several months. Some of the cases in the U.S., and how they’ve appeared are referred to as “cryptic transmission,” which basically refers to “undetected” transmission. But we do know how to contain pathogens, and precautions including principally distancing from the ill or potentially-ill, hand washing, masks, etc. have been proven to be very effective.
Atul Gawande has recently written about his experiences in implementing some of Hong Kong’s and Singapore’s best practices in dealing with SARS-CoV-2—interventions that have led to zero transmission in some clinics, even with positive cases presenting on a daily basis.6 Fear will prove to be a worse adversary than this strain of coronavirus. Cases of coronavirus cannot spread to where there is no viral vector. Remember, Louis Pasteur conducted groundbreaking experiments on the propagation of microorganisms which led to the germ theory of disease, and in 1864 he announced “La génération spontanée est une chimère,” (‘The spontaneous generation [of microbes] is a fantasy’ [author’s translation]).
In positive news that falls in the ‘Win’ column for science, Gilead’s antiviral therapy, remdesivir, originally trialed for Ebola (where it was unsuccessful), is being re-purposed in trials for coronavirus, because of the way this drug works on the polymerase of the viral genetic code to prevent proper re-assembly within host cells. There are dozens of compassionate use cases and trials ongoing for this treatment, and there are more data to follow. Demand has now been so high, outstripping supply, that Gilead has had to suspend availability outside of the clinical trials until more is known about its performance and the crucial balance of benefit versus risk.
Additionally, on March 16, we dosed the first human patient with mRNA-1273, which is Moderna’s coronavirus vaccine candidate. This clinical trial is likely to last until about November, so for the foreseeable future, we’re going to have to rely on emerging public immunity and social distancing protocols to slow and abate the spread. In the meantime, news stories will continue to snowball, as news stories are want to do. There have been various reports that there are two different strains of SARS-CoV-2, which are also dually-purported by doomsayers to ensure the apocalypse. I can tell you, having personally reviewed the genetic codes for both strain candidates—currently known as Type L and Type S—that there are still arguments as to whether the differences represent appreciably-different strains. The initial stratification of 103 cases suggested that 70% of them were the L-type, and ~30% were the ancestral S-type lineage. The authors of this nascent research suggested that selective pressure was at the root of this split, which is unsurprising for viral reassortment. They note, “Human intervention may have placed more severe selective pressure on the L type, which might be more aggressive and spread more quickly. On the other hand, the S type, which is evolutionarily older and less aggressive, might have increased in relative frequency due to relatively weaker selective pressure.”5
The RNA code for SARS-CoV-2 is about 30,000 letters long, and it’s suspected (by Bedford’s team) that it changes by about 1 letter every 15 days.7 So, for it to be substantially different through genetic shifting or drifting is expected to take quite some time. It needs enough ‘letters’ to change to look different ‘enough’ to the immune system to constitute a new infection.
Herd Immunity and Flattening The Curve
I’ll close with two powerful ideas here: Herd Immunity and Flattening The Curve (see Figures 2 and 3). For those who have had COVID-19 and have recovered well, they will have immunity, and will begin to provide the herd immunity in public to prevent new and future cases from progressing from infected people through vectors to naïve patients. This is exactly how immunizations are intended to work, such as with vaccines for influenza or measles. It requires that a substantial proportion of the population be immunized, and then new vectors are very difficult to sustain. There’s community spread of SARS-CoV-2, and the viral vectors appear to be very full. Likely the only resolution to the pandemic now will be to sustain adequate numbers of those immune in the population so that we see the vectors breaking.
The idea of “flattening the curve” is so that we don’t have a simultaneous crushing demand on the nation’s 100,000 ICU beds (Johns Hopkins’ estimate), nor the 925,000 staffed hospital beds. What we’ve learned, epidemiologically, from the 1918 pandemic flu is that spreading out infections over time—compared with simultaneously experiencing them all at once—lowers the overall death rate (even though it will make the pathogen circulate longer). So, by slowing the spread, we stand a chance of keeping up with the critical need patients that require ventilators and other emergent interventions. And much like I last suggested in the March issue, upwards of 80% of cases will still be mild, and not require medical care.
In closing this feature, I’d like to say ‘good luck out there,’ but luck doesn’t apply in #science. Instead, do what we know to be effective. Stay out of groups and where congregating people share a volume of air—this is primarily a respiratory-transmitted pathogen, even though people have gone crazy cleaning surfaces. Don’t touch your face, and keeping 3 meters from others is more effective than 1 or 2 meters.
- Taleb, N.N. (2009). The Black Swan: The Impact of the Highly Improbable.
- Pasteur, L. (1878). Les Microbes organisés, leur rôle dans la Fermentation, la Putréfaction et la Contagion.
- Zhuang, G.H. et al. (2020). Potential False-Positive Rate Among the ‘Asymptomatic Infected Individuals’ in Close Contacts of COVID-19 Patients. Zhonghua Liu Xing Bing Xue Za Zhi. 41(4):485–488. doi:10.3760/cma.j.cn112338-20200221-00144
- Ioannidis, J. (2020). A fiasco in the making? As the coronavirus pandemic takes hold, we are making decisions without reliable data. STAT.
- Xiaolu Tang, Changcheng Wu, Xiang Li, Yuhe Song, Xinmin Yao, Xinkai Wu, Yuange Duan, Hong Zhang, Yirong Wang, Zhaohui Qian, Jie Cui, Jian Lu, On the origin and continuing evolution of SARS-CoV-2, National Science Review, nwaa036, https://doi.org/10.1093/nsr/nwaa036
- Gawande, A. (2020). Keeping the Coronavirus from Infecting Health-Care Workers. The New Yorker.
- Johnson, C.K. (2020). Scientist links 2 state outbreaks with genetic fingerprints. ABCNews.
Ben Locwin, a healthcare executive, has worked on vaccinology and virology advisory boards, and has been on the front lines of the national coronavirus pandemic, assessing patient cases, local and national epidemiology, and providing recommendations for government and private sector response amid its continued impact to public health and world economies.