Read on Twitter
How many people in the US have COVID?
We don’t know, but @inschool4life and I estimate how many people have gone to their doctor with a non-flu influenza-like illness (ILI) in March.
https://github.com/jsilve24/ili_surge/blob/master/Silverman_and_Washburne.pdf
We don’t know, but @inschool4life and I estimate how many people have gone to their doctor with a non-flu influenza-like illness (ILI) in March.
https://github.com/jsilve24/ili_surge/blob/master/Silverman_and_Washburne.pdf
Using data from @CDC, our estimates suggest a surge of over 9 million people visited a doctor with a non-flu ILI the week of March 15-21.
This is well above the baseline, and states w/ more COVID are seeing a larger surge.
This is well above the baseline, and states w/ more COVID are seeing a larger surge.
These non-flu ILI surges are correlated with COVID case counts across states, a correlation that grows stronger with time.
This suggests that COVID could be a significant fraction of these 9 million outpatients.
This suggests that COVID could be a significant fraction of these 9 million outpatients.
The COVID case counts are much smaller than the ILI surges. If these surges are comprised of mostly patients with COVID, it suggests US COVID testing may only be finding 0.75% of COVID cases.
9 million isn’t an unreasonable number.
W/ onset of community transmission Jan 15, a 3 day doubling time (dt) predicts 9 million cases. If the dt in the US has been < 3 d, it could explain the 9mil cases with a higher subclinical rate.
W/ onset of community transmission Jan 15, a 3 day doubling time (dt) predicts 9 million cases. If the dt in the US has been < 3 d, it could explain the 9mil cases with a higher subclinical rate.
What doubling times do we observe in the US?
Both case counts and COVID deaths across the US suggest the US epidemic could be growing faster than a 3-day doubling time would suggest.
Both case counts and COVID deaths across the US suggest the US epidemic could be growing faster than a 3-day doubling time would suggest.
EMPHASIS: We don’t know the exact number of COVID cases in the US. We see a surge of non-flu ILI patients. This could be from changing rates of ILI presentation due to widespread fear of COVID or from another coincidental non-COVID, non-flu ILI.
Our work has many limitations. We are not making recommendations on public health interventions, personal infectious disease risks or other decisions with life or death consequences.
With more data, our analyses make predictions that can be tested. Serosurveys can test if non-flu ILI patients from March had COVID at the rate we predict here. Serosurveys can also reject our hypothesis.
SCIENCE!!!
SCIENCE!!!
*IF* serosurveys corroborate ILI surges as estimators of the prevalence of emerging infectious diseases, then other syndromic datasets of common clinical presentations may assist the next pandemic, which may not present like flu.
