Read on Twitter
Dr. @sdbaral has an excellent guiding philosophy here.
As someone who entered science to curiously question the answers, I've found epidemiology during COVID to be characterized by strong informal social control to reduce appearance of scientific uncertainty.
For example... https://twitter.com/sdbaral/status/1325063454139551744
As someone who entered science to curiously question the answers, I've found epidemiology during COVID to be characterized by strong informal social control to reduce appearance of scientific uncertainty.
For example... https://twitter.com/sdbaral/status/1325063454139551744
... there's reasons to revisit and reassess our estimates of population-level severity and, with it, individual severity of COVID. The discovery of T-cell positive, seronegative patients illuminates a larger pool of unascertained cases than previously estimated by serosurveys.
For me, the early studies of COVID timeseries produced what I called an "accounting gap". The first thing I noticed was that growth of cases in early epidemics was faster than most estimated. When you combine this with early start-dates, it points to a curious mystery...
If the US epidemic started January 15th and doubled every 2.5-3 days, then it would produce an enormous number of patients. The same goes for China: with a mid-November start date and 2-3 day doubling times it's mathematically possible for all of China to have been infected...
Evidence for the unmitigated nature of a Chinese epidemic comes from what Levitt published & I witnessed independently: a regular decay in the growth rates of cases/deaths in Hubei province prior to lockdowns there.
We can all agree: China was not transparent in this epidemic.
We can all agree: China was not transparent in this epidemic.
When they reported 10 cases a week in Guangdong Province or Hong Kong, they were passing laws to seize private property for massive quarantine facilities there.
This skepticism of China's epidemic translates to our analysis and interpretation of the rest of the world's epidemics
This skepticism of China's epidemic translates to our analysis and interpretation of the rest of the world's epidemics
For starters, a 1% infection fatality rate in China would imply nearly 8.4 million deaths, most unreported to the WHO. China wanted to hide 1-3 million Uighurs in a concentration camp, so it's not unreasonable...
However, the early start-date and fast growth rate in the US produced an "accounting gap" of too few arrivals to the doctor measured through ILI visits to outpatient providers than one predicts from a 1% infection fatality rate.
This early skepticism about China was what got me looking at ILI data in February to look for a surge in densely populated and internationally connected states and assess if that surge is consistent with a fast-growth/lower-severity hypothesis.
https://stm.sciencemag.org/content/12/554/eabc1126
https://stm.sciencemag.org/content/12/554/eabc1126
The paper by Sunetra Gupta and Jose Lourenco pointed out similar uncertainty in the UK - in the UK, the fast growth combined with an *unknown* start date to allow a range of possible explanations that vary primarily in their subclinical rate. https://www.medrxiv.org/content/10.1101/2020.03.24.20042291v1
Notice in the title of their paper: "the immediate need for serosurveys". Serosurveys would help us determine the ascertainment rate of cases, the subclinical rate, where reality fell in this accounting gap between fundamental principles of epidemic spread and our observed cases.
Serosurveys then provided some insights: Gangelt found a 0.4% IFR from their serosurvey; New York 1% IFR (this shocked many: some said our estimates of 10% of New York State were outrageous, but they found 13% seroprevalence just weeks later.
(admittedly: our estimates did change with peer review. Our first approach to estimate prevalence from ILI were correctly pointed out to be too high by a very sharp-penciled analysis from a Twitter user - everyone is on team science, and we were grateful for the feedback!).
Serosurveys filled in some of the gap between models of epidemic spread based on fast-growth and early start-dates and the reality of patients counted.
But they weren't quite there yet: it's possible some patients didn't seroconvert, or their seropositivity waned.
But they weren't quite there yet: it's possible some patients didn't seroconvert, or their seropositivity waned.
We pointed out this uncertainty in terms of an unknown "subclinical rate".
If we assume 0% subclinical rate (purple line) and a 16 day lag from ILI to death, then we get a ~1% IFR.
If we assume 33% subclinical rate (magenta), similar to flu, we get a 0.3% IFR.
If we assume 0% subclinical rate (purple line) and a 16 day lag from ILI to death, then we get a ~1% IFR.
If we assume 33% subclinical rate (magenta), similar to flu, we get a 0.3% IFR.
T-cell surveys are finding many patients that didn't seroconvert but do have evidence of SARS-CoV-2 infection by specific recognition of SARS-CoV-2. However, n is small in these studies, so there are huge error bars about how much we should correct for this.
If COVID had a 0.3% IFR - or perhaps less on average after the first wave in March, as vulnerable patients from April to present may have been better able to protect themselves (see figure below for changing % positive by age) - then 0.1% of US deceased implies 1/3 infected.
This unknown infection fatality rate - and unknown attack rate at herd immunity - combine to produce an observed range of possibilities for deaths per-capita at herd immunity.
I've been following this track since April, sharing results with some, but withholding a paper...
I've been following this track since April, sharing results with some, but withholding a paper...
The main reason for withholding the paper is that I'm afraid of flak from loud scientists and citizens if I go against the flow with this hypothesis. Below is tip-of-the-iceberg claims of how such papers are "irresponsible" and I'm "responsible for deaths".
However, there are scientific reasons, too: some very real and important synchronized changes in case growth rates across US states in mid-August materially changed the results. The paper predicted summer waves neatly, some adjustments were needed for fall/winter waves.
I'll start finishing up this paper and try to get it in a reasonable and humble format. Still waiting for feedback from @VPrasadMDMPH @WesPegden @BallouxFrancois and others on the rough draft. I'm super grateful for the zen-master advice on "the flow" from @sdbaral.
I'm sorry for any/every time my words have felt unkind. Some of that has been me pushing back against what felt like a barrage of hostility from the institution I loved for all of my life before 2020 (and whose ideals I still love).
I'll try to get this science out, kindly.
I'll try to get this science out, kindly.
I'm super grateful for all the people who've listened to and provided feedback on this work in private. I'm also grateful to people who've carved out space for scientists to keep pursuing reasonable curiosity, wherever that may lead.
I love you all. Science!!!
I love you all. Science!!!
"In the end, we remember not the voices of our enemies, but the silence of our friends."
These words are so true. As disappointed as I am in the people I thought were friends who left me to be eaten by the piranhas, I've also found true friends in battle. You're all the best!!!
These words are so true. As disappointed as I am in the people I thought were friends who left me to be eaten by the piranhas, I've also found true friends in battle. You're all the best!!!
Stay healthy. Do your best to reduce transmission until we know this is over. Listen to doctors and nurses - they are overwhelmed in many places and need a helping hand. Be nice to people, especially those you disagree with.
And please, help us tell the nuanced history of COVID.
And please, help us tell the nuanced history of COVID.
