Early in the Covid pandemic, Michael Levitt noticed a gradual decay of case growth rates over time in Wuhan, and many dismissed or ignored his observations on account of what they viewed were improper credentials and unconventional mathematical methods (Gompertz curves, as opposed to conventional compartmental models in epidemiology).
Some researchers went so far as to call Michael Levitt’s work “lethal nonsense,” saying he was being an irresponsible member of the scientific community by not being an epidemiologist and presenting work that Levitt’s critics believed downplayed the coronavirus.
On March 17, 2020, John Ioannidis argued that Covid severity was uncertain and extreme containment policies such as lockdowns could possibly cause more harm than the pandemic itself, provoking a persistent culture of animosity towards Dr. Ioannidis, from false claims of conflicts-of-interest in 2020 to people accusing Ioannidis of “horrible science” and more.
My Experience as a “Deviant” Epidemiologist
As a mathematical biologist studying viruses jumping from bats to people for a few years prior to Covid, and as a time-series analyst with nearly a decade of experience forecasting by early 2020, I was also studying Covid since January 2020.
I noticed the wisdom of Levitt’s Gompertz curves – Levitt found an observation I myself had found independently, of regular decays in the growth rate of cases well before cases peaked in Wuhan, and then in early outbreaks across Europe and the US. In my own work, I found evidence in February 2020 that cases were doubling every 2-3 days (midpoint estimate 2.4 days) in the early Wuhan outbreak at a time when popular epidemiologists believed Covid prevalence would double every 6.2 days.
We knew at the time that the earliest cases were exposed in late-November 2019. Suppose the first case was December 1, 2019, 72 days prior the approximate early-2020 peak of cases in China on February 11, 2020. If cases strictly doubled every 2.4 days over that 72-day period, as many as 1 billion people, or 2/3 of China, would have been infected. If, instead, cases doubled every 5 days, we’d expect roughly 22,000 people to be infected in China.
If cases doubled every 6.2 days, we’d expect 3,100 people to be infected in China. The slower the case growth rate one believed, the fewer cases they expected, the higher the infection fatality rate they estimated and the more severe they worried the Covid-19 pandemic would be. These findings led me to see the merit in Dr. Levitt’s observations, and to agree with Dr. Ioannidis’ articulation of the scientific uncertainty surrounding the severity of the Covid pandemic the world was about to experience.
However, when I saw the world’s treatment of Levitt, Ioannidis, and many more scientists with contrary views that mirrored my own, I became fearful of possible reputational and professional risks from sharing my science. I tried to share my work privately but encountered professors claiming I was “not-an-epidemiologist”, and one told me I “would be directly responsible for the deaths of millions” if I published my work, was wrong, and inspired complacency in people who died of COVID.
Between these personal encounters from scientists in a variety of positions and the public stoning of Levitt and Ioannidis, I worried that publishing my results would result in me being publicly called not-an-epidemiologist like Levitt, and responsible for deaths like both Levitt and Ioannidis.
I managed to share my work on a CDC forecasting call on March 9th, 2020. I presented how I estimated these fast growth rates, their implications for interpreting the early outbreak in China, and their implications for the current state of COVID in the US. Community transmission of Covid in the US was known at the time to have started January 15th at the latest,
I showed how an outbreak starting in mid-January and doubling every 2.4 days could cause tens of millions of cases by mid-March, 2020. The host of the call, Alessandro Vespignani, claimed he didn’t believe it, that the fast growth rates might just be attributable to increasing rates of case-ascertainment, and ended the call.
Just 9 days after I presented on the CDC call, it was found that Covid admissions to ICUs were doubling every 2 days across health care providers in New York City. While case-ascertainment could be increasing, the criteria for ICU admission, such as quantitative thresholds of blood-oxygen concentrations, were fixed and so the ICU surge of NYC revealed a true surge of prevalence doubling every 2 days in the largest US metro area.
By late March, we estimated an excess of 8.7 million people across the US visited an outpatient provider with influenza-like illness *ILI) and tested negative for the flu, and this estimate of many patients in March corroborated a lower estimate of COVID pandemic severity.
Having watched Levitt, Ioannidis, Gupta and more get mobbed online for publishing their evidence, analyses and reasonings for a lower-severity pandemic, I knew that publishing the ILI paper was an act of deviance in an extremely active online scientific community. My motivation wasn’t to be a deviant, but to carefully and accurately estimate the number of people infected, and to present these estimates to the world, because the world needed to know how bad COVID would be to react proportionately to this novel virus.
However, after we released the ILI paper on the preprint server, the paper got picked up by a brilliant team of data journalists at the Economist and went viral. As the paper went viral, the onslaught reputational and professional threats I’d feared began to materialize.
Colleagues said I risked being “responsible for the deaths of millions” (a crime on par with genocide, if the comment is taken literally), that I had blood on my hands, that I was “disrupting the public health message,” that I was “not an epidemiologist,” and more. The verbal stones came from all sides, from people who were once colleagues and friends to members of the scientific community I’d never heard of before saying I killed thousands.
The Science Not Shared
I continued to study this alternative theory of Covid based on faster-growth and its implied lower-severity. Under this theory, it’s possible New York City reached herd immunity in its March 2020 wave and, if so, then features of the outbreak in New York City could be used to predict outcomes from later uncontained and lesser-mitigated outbreaks in places like Sweden, South Dakota, and Florida.
I estimated Covid cases in the Fall 2020 outbreaks would peak around 1 death per 1,000 capita or 340,000 deaths. At that time, prominent epidemiologists whose views aligned with “the message” were still using estimates of high-severity outcomes, where millions of US deaths would be possible if the virus went uncontained.
However, having experienced the barrage of hostilities leading up to and after the ILI paper, and seeing the continuation of hostilities towards a rotating cast of scientists with similar findings that deviated from “the message,” I worried about sharing this full theory.
I watched carefully in the summer 2020 as the unexpectedly low and early peak of cases in Sweden baffled epidemiologists but aligned neatly with my theory. I watched the Fall 2020 outbreaks from Chicago to South Dakota decelerate, as Levitt had noticed, and peak earlier than we’d expect from seasonal forcing and in a manner consistent with the Mar-April 2020 NYC outbreak. The median US county peaked around 1 death per 1,000 capita, the US outbreak peaked at around 350,000 deaths, and outbreaks in hundreds of relatively unmitigated counties saw cases decline before the arrival of vaccines.
I eventually released these forecasts and findings in April 2021, after vaccines had enough time to roll out and hopefully then nobody would claim I was disrupting “the message.” I deliberately withheld these findings from preprint servers due to a justified fear of hostility from the scientific community during COVID-19.
By creating a research environment hostile to evidence of a lower-severity pandemic, the science people read on the news to inform their beliefs and actions of overestimated Covid risk. That science was not the result of a fair competition of ideas won by evidence and logic, but a silencing of ideas by federal officials coordinating devastating takedowns of competing views, by biased social/mass-media amplification of one theory, and by a norm of private and public hostilities enforcing a particular theory of Covid-19.
Informal Censorship of Science in COVID-19
Censorship takes many forms. The most extreme form of censorship is the formal criminalization of speech, such as the arrests of people in Russia protesting Putin’s war on Ukraine.
Science in Covid-19 was not censored through any formal social control such as laws prohibiting speech or publication of particular results. Science was, however, silenced by informal social control, by scientists in our community enforcing, by words and deeds, a narrow range of scientific beliefs and unscientific norms and values about who could present a scientific finding or theory, or who can make a unique point without being harassed by colleagues.
Whether attacking Levitt and Ioannidis or the Great Barrington Declaration signers Jay Bhattacharya, Martin Kulldorff and Sunetra Gupta, scientists utilized social media platforms and mainstream-media outlets to take down competing views from other scientists. But Washington Post, BuzzFeed, or New York Times articles are not venues to resolve scientific uncertainty or advance scientific debates; they’re venues to amplify a message, and the message being amplified was that estimating COVID risk to be lower than a clique of epidemiologists is wrong or immoral and should not be considered or is not relevant when discussing pandemic policy.
Twitter, a war zone well-known to amplify incendiary content, is not the place to resolve scientific debates, but it’s commonly a place to call people out and mobilize angry mobs capable of getting people fired.
The public attacks of scientists were attempts at public executions, and we humans have a long and troubled history of public execution. Historically, public executions were believed to better deter deviance from laws and authorities, and the public punishments in Covid served a similar purpose of discouraging onlookers like me from doing anything that could be remotely interpreted as similar to the crime that got great Stanford scientists stoned.
The sociological effect, and quite possibly the intent, of attempted public executions of scientists highlighting uncertainty in Covid outcomes or, worse yet, estimating lower severity of Covid pandemic burden, was the informal social control of scientists like me who analyzed Covid-19 data every day of 2020 and sat on findings highlighting uncertainty or estimating lower severity.
In criminology, social control theory attempts to explain why some people commit crimes and others do not, and I find social control theory is most useful to understand my own choices to not publicize my work in mid to late 2020.
Throughout 2020, I witnessed how social media platforms and mass-media became tools to manufacture the consent of the public to agree with a powerful clique of epidemiologists. These epidemiologists claimed their science was uncontested and protected their scientific theories from contest by public broadcasting of sanctions against fellow scientists. Shame, criticism, ridicule, disapproval, and other checks on deviance from norms and values of publishing work in agreement with this clique of epidemiologists, or from experts they approve of.
Such informal social control on scientific findings has no place in any reasonable ideal of science in a society. If we allow scientists to take down other scientists through personal attacks, if we fail to disentangle a complex of close associations between scientists and the mass media they use to manufacture belief in their own theories, then what we call “science” would be battle over belief mediated not through the peaceful and cooperative ideals of evidence and reason, but by the savage violence of cultural warfare. It becomes a barbaric media battle to achieve scientific dominance by ridiculing dissidents and suppressing dissent through informal social control.
A Path Forward
If, however, we unsparingly examine the use of media in science, and the practice of high-profile attempted public executions by famous scientists, we can identify a sociological cancer in our science and eradicate it before it metastasizes any further. The science we never share risks being a finding we never found.
As the pile of unshared science grows, our scientific understanding of crises like pandemics suffers from the attrition of the science it doesn’t know. It should be in the interest of all scientists to facilitate the sharing of scientific ideas to ensure no science goes unshared from fear of ridicule or public execution.
Thankfully, we are scientists. We can innovate new platforms and institutions, and create better and more professional media for the exchange of scientific ideas, we can reform science before the next pandemic.