Outer Limits of Plausibility: Omicron IFR
Detailed look at death data and what it means for Omicron's IFR
An infection fatality rate (IFR) is often estimated rather than computed from hard data, because it is made from the ratio of ‘deaths/infections’ — and the bottom number (the total infections) is rarely known with 100% confidence.
An IFR estimate has a natural lower bound but not a natural upper bound. The natural lower bound is the innate lethality of the disease — and the outcomes from having the disease don’t get any less lethal than that.
But outcomes can be worse than the worst-ever outcomes that the disease itself is capable of causing. This is because sub-optimal medical treatments can be used, leading to even higher death than the disease, itself, would have caused.
With IFR estimates that are only bounded on one side like that, the correct way to estimate the IFR is to give more weight to the lower ones, because higher IFRs are not naturally bounded, and they can rise and rise, depending on whether good medical treatments got employed or not.
Even something so simple as a paper-cut, if treated very haphazardly — such as by rubbing dirt into it — can lead to outcomes worse than those paper-cuts would EVER be able to cause, if they were treated well (with proper infection control).
The ideal case for estimating an IFR then, is a large study finding a low IFR, because it can be assumed that the disease was treated well — and that the IFR which did arise, arose from the inherent lethality of the disease itself, and not from improper intervention.
From that ideal circumstance (large study, low IFR) you can add an interval estimate to get out to even the outer limits of physical plausibility. A good interval estimate is a 99% credible interval, because it has only a 0.5% chance of not containing a true IFR which is higher.
For even more certainty, however, a 99.99% credible interval could be used, so that there is only a 0.01% chance of the interval not capturing the true value, and only a 0.005% chance for the actual, underlying IFR to be higher.
That’s just one chance in 20,000 — and it is far beyond the confidence level typically utilized when presenting scientific evidence purportedly showing that a new medicine “works.”
Here is a 99.99% credible interval on the IFR for Omicron variant, as put through the statistics software, R:
Even the 99.99% upper bound IFR on Omicron is below what is seen with seasonal flu, when the average of the 7 flu seasons spanning from Fall of 2012 up to Spring of 2019 are used for comparison.
The most-likely IFR for Omicron is just 0.035%, which makes it just under one-third as lethal as seasonal flu.
This point estimate, where seasonal flu is 3 times more deadly than Omicron, comes from the average of all of over 2000 random IFRs which happened to have led to results seen in an Ontario study — when over 37,000 people got Omicron, and 12 died.
If other and higher estimates of Omicron IFR get published, keep in mind that there is no natural upper bound for estimation, and then “down-weight” those studies based on the presumption that cases were either not given acceptable standard-of-care, or the population under study, itself, was less than healthy.
Omicron IFRs greater than, say, 0.09%, contradict known data (they are not consistent with known evidence).
Try as you might, you will not be able to replicate the Ontario Omicron death data with an IFR above 0.09%* — so you could go so far as to say that an IFR which is that high actually contradicts reality (the telltale sign of a falsehood, even if put forward as a truth).
Physical limitations prevent water from boiling at 140 degrees Fahrenheit (at least in an air pressure high enough to be breathable) and, by the very same reasoning, biophysical limitations prevent Omicron IFRs from being above 0.09%.
*A million trials with IFR above 0.09% may yield a few trials which replicate the Ontario data, but “a-few-in-a-million” isn’t saying much.
Reference
[12 total deaths from 37,296 Omicron infections] — Ulloa AC, Buchan SA, Daneman N, Brown KA. Estimates of SARS-CoV-2 Omicron Variant Severity in Ontario, Canada. JAMA. Published online February 17, 2022. doi:10.1001/jama.2022.2274. Available: https://jamanetwork.com/journals/jama/fullarticle/2789408