Question: A lot of folks are reading a post on Medium yesterday that shows that we’re not being data driven and we’re all overreacting to coronavirus. Did you read it? What do you think?
Answer: I did read that post on Medium, which has since been taken down because it is “under investigation or found in violation of Medium’s rules.” The author, Aaron Ginn, a Silicon Valley self-described “technologist” who has written for Breitbart, re-posted to a website called ZeroHedge. In short, this post was exceptionally problematic and I’m glad that Medium has taken it down because it is such a flawed reading of data that gives readers incorrect information and the wrong take-aways. Rather than just leaving it at that, I feel obliged to walk through some of the overarching problems and I am going to do so by going though each of the post’s subheadings. Here’s my peer review of Andrew Ginn’s post.
Ginn’s overarching theme — We are all overreacting to the coronavirus and our overreaction is worse than the virus itself.
Response: I am angry too, but for a different reason. The United States did not get testing rolled out quickly enough — both for testing suspected cases and for conducting ongoing population-based surveillance — to keep the virus from moving through our communities. And because we missed that boat, we are all now gravely suffering an economic catastrophe. Even though we were late, our current rush to contain the virus is warranted. Listen to the experts and not to this self-described “technologist.”
Response by subheading:
1) Total cases are the wrong metric: Fair point. But no one is pointing to total cases in isolation, even Johns Hopkins highlights total cases alongside total deaths and total recovered. And most of the news coming out of places like China is focused on new case counts per day. Furthermore, understanding the number of people a new virus has infected is not a “vanity data point” as the author points out. It’s our denominator for all other analyses! Want to talk about case mortality rates? First, we need to know the number of cases.
2) Time lapsing new cases gives us perspective: Yes! And as a reminder, whenever you’re looking at a graph, first look at the X and Y axis. Here, the X axis is days, in 5-day chunks. But the Y axis, it’s different — it’s a logarithmic scale. We use log scale when we’re dealing with orders of magnitude (e.g. the values at the higher end are much larger than the values at the lower end of the range). So, what do we see — the case count growing at exponential rate. What’s the take-away? Answer: A small outbreak can quickly become a massive outbreak.
3) On a per capita basis, we shouldn’t be panicking: Agree. Panic is not helpful. And yes, it’s good to put things into perspective, so I appreciate the per capita breakdown. But remember the above section? It means that individual risk can change rapidly.
4) COVID-19 is spreading, but probably not accelerating: Even if the rate of spread remains constant, as long as the rate is greater than 1, we will see acceleration in the daily numbers of people who contract the virus. But let’s look further at what the author posits. He states, “daily growth rates declined over time across all countries regardless of particular policy solutions, such as shutting the borders or social distancing.” And then shares data from China, South Korea, Italy, and Iran. These countries have all implemented stringent controls over time, so to state that growth rates naturally decline is a fallacy. There’s no counterfactual here to test against. Second, look at the title of the graph, it’s “daily percentage increase in #of reported COVID-19 cases.” Daily percentage increase in # of cases is NOT equivalent to daily growth rates. If we start with 40 cases on day 1 and have a 200% increase, it’s an increase of 80 people, bringing case count to 120. Now if those new 80 people infect another 160 people, our case count on day 3 will be 280, but that’s only a 43% increase (we had a 200% increase the day before!). Why am I bringing this up? To point out that the rate of spread is difficult to visualize in charts like this.
5) Watch the bell curve: I’m honestly not sure what this section of the paper is trying to convey. The author states that “CDC and WHO are optimizing virality and healthcare utilization, while ignoring economic shock to our system. Both organizations assume you are going to get infected, eventually, and it won’t be that bad.” Huh? Where is this massive assumption on behalf of the author coming from? And taken together, these sentences don’t make sense. The first sentence is that CDC and WHO are trying to save our health system and second is that CDC and WHO don’t think it will be that bad if a person gets infected. And when it comes to exponential growth of any given thing, when there are limiting factors (here, number of human hosts for the virus), yes, you will generally have a bell curve. So what?
6) A low probability of catching COVID–19: This all depends on how quickly we can control the pandemic. The wider the pandemic expands, the greater your probability of catching the coronavirus. Data from Wuhan (see Lancet study here) show that before strict measures were enforced, the daily reproduction number (R0) was 2.35 and declined to 1.05 after travel restrictions were introduced. Government policy and collective action of individuals can have an impact on spread of the virus. Indeed, the same WHO report the author cites to boost his dubious claim that “research show that COVID-19 doesn’t spread as easily as we first thought or the media had us believe” states:
“The COVID-19 virus is a new pathogen that is highly contagious, can spread quickly, and must be considered capable of causing enormous health, economic and societal impacts in any setting. It is not SARS and it is not influenza. Building scenarios and strategies only on the basis of well-known pathogens risks failing to exploit all possible measures to slow transmission of the COVID-19 virus, reduce disease and save lives.”
A couple more points — we know that when it comes to respiratory viruses, mass gatherings can contribute to case spread and we have evidence from various recent experiences to support this claim — see, for example, the connection between Boston COVID-19 outbreak and the Biogen Conference. And we know that with exponential growth, the risk to the population grows concurrently. A low risk today does not equate to a low risk next week.
7) Common transmission surfaces: Agree, this study cited shows that the virus can live on surfaces, but did not go so far to show whether the virus can be transmitted from these surfaces to individuals. The key take-away on this front is that we absolutely have to keep washing our hands and avoiding touching our faces and cleaning our spaces because fomite transmission is within the realm of possibilities and now we have more data showing that it’s plausible.
8) COVID-19 will likely “burn off” in the summer: We have no evidence to support this claim. We can hope it’s right, but let’s not incorrectly use data to support this hope. First, we do have a number of cases in the southern hemisphere now! Check out data from Australia, Brazil, Malaysia, Singapore, etc. Second, many countries in Africa do not have widespread testing — we shouldn’t confuse the lack of data with the lack of cases. Third, this Harvard Center for Communicable Disease scientist, Dr. Lipsitch, puts it best, when answering the question, “Will COVID-19 go away on its own in warmer weather?”. His answer, “probably not.” To understand his rationale and the science behind it, check out his post.
9) Children and teens aren’t at risk. Yes, the good news in this pandemic is that children and teens seem to fair much better than adults, especially older adults. Children and teens become infected at similar rates as adults, and it’s still unclear how much transmission from children to children or children to adults occurs. Now, there is a risk difference between babies, children <10 and teens. We have seen some adverse outcomes in babies, some death in teens and no deaths so far in children. So, risk of adverse outcomes increases with age. And risk to teens is much, much less than risk to adults, especially older adults. But claiming no risk is also incorrect. This recent paper in Pediatricsshows that there were severe and critical cases among children and teens infected with COVID-19 in China. So, let’s not make sweeping claims that are incorrect in an effort to assuage people’s fears.
10) A strong, but unknown viral effect: The degree of coronavirus spread will depend on the interventions we implement. We can change R0. Indeed, the change in R0 in China was not a result of natural shift; it was the result of China’s extremely strong efforts to curb the virus’s spread. What we’ve seen in places that are at the beginning of the epidemic and places that have not implemented strong measure to control the epidemic is a growth rate that is exponential — with each person on average infecting at least 2 others.
11) What about asymptomatic spread?: The author states that data on asymptomatic spread “is still unclear but increasingly unlikely.” It’s actually the reverse. Yes, asymptomatic spread is still unclear, but the data is pointing to it being increasingly likely. Nature had a good summation of the evidence to date in an article published on 3/20. This evidence further points to the need for social distancing measures to be in effect.
12) 93% of people who think they are positive are not: This is a misuse of data. First, getting a test does not mean a person thinks they are positive. Second, the testing strategies have varied dramatically by country and even over time in country, percent positive in the US now and percent positive in South Korea have no relationship without that context. Most importantly, widespread testing is NOT about understanding a person’s underlying risk of acquiring coronavirus. Rather, it’s about identifying who is infected so that we can track their close contacts, minimize spread, protect health workers who may come in contact with the infected, and identify treatment strategies, if possible.
13) 1% of cases will be severe: No. We are still understanding mortality rates connected with COVID-19 and using US data from mid-March, before the US was more widely testing and before most cases had been resolved is completely inappropriate. As I’m typing, the US has 26,747 cases and 340 deaths (per Hopkins), which yields a very crude case fatality rate of 1.2% and that’s not including all the other cases that are severe but don’t result in death. Moreover, if our health system falters as we’re seeing in Italy, we could have much higher case fatality rate (it’s 9% there right now!).
14) Declining fatality rate: Yes, as we capture more mild cases of COVID-19 through expanded testing, we can hope to see our fatality rate decline. And as we protect our health system so that it can adequately respond, we should also hope to see more people survive. If we do not do the needful, however, we will see increasing fatality rates, as Italy is seeing now. China had declining fatality rates because of strong interventions, not by some magic.
15) What should we do?: Stop spreading false narratives. Stop basing decisions on faulty read of the data. Follow the advice of public health experts. Ramp up testing NOW. Hold our leaders accountable for inaction earlier when it could have staved off this health and economic crisis. Bring economic relief to the millions of Americans suffering.
In case you’re wondering who I am and what credentials I bring to this peer review:
I am a trained public health professional. I received my Master in Health Science from Johns Hopkins Bloomberg School of Public Health, where I studied demography and epidemiology. For the last 13 years, I have worked as a Public Health Advisor first for the CDC and most recently for the U.S. Agency for International Development. I have written peer-reviewed publications published in journals including Bulletin of the WHO and Lancet. I serve on the Editorial Board of Global Health Science and Practice and have served as peer reviewer for a variety of public health journals. The views shared herein are my own and do not represent the views of the U.S. Government or any other entity with which I am affiliated.