Everyone recalls the fear from coronavirus era, and this website was amongst the earliest to identify the statistical pattern consistent with pandemic [versus “just the flu”], and warn about it at the start of 2020 before the global traction took hold. Steeped in both statistical anomalies and literacy of the patterns in the Wuhan 1.0 cases, and the asymptoticly slow, government policy response was a key to it’s foretell. Now the news media exploits the news cycle with images of patient cases among global ship travelers now, disembarked or quarantined travelers, and frighteningly high fatality rates. Enough to create the psychological whiplash to the Covid-19 era. But there are some critical differences and their global potential, and the statistical data so far reveal much to be learned from both viruses.
what is hantavirus?
This is
primarily rodent-borne, aerosolized urine, droppings, or feces. Unlike coronavirus, the transmission is rare
between humans, and it is right now far limited to the Andes-strain. Or from Argentina travel.
The case
fatality ratio is what makes this virus more striking than most other global
diseases in our recent past. While the
virus has known to be existent for a small number of years, it is quite deadly. At up to 50% death rate in the Americas,
though currently lower [but still deadly] elsewhere.
Now while
the fatality rate is high, it is important to note the transmissibility of hantavirus
is very low. So for example, while coronavirus
had a high transmission rate [R₀ >> 1] allowing for convex growth that
overwhelmed systems, hantavirus has a R₀ ≈ 0 in almost all settings.
The copula
mix of two dependent variables is therefore less optimal. The virus behaves more like a localized trap
from zoonotic vectors, rather than a global outburst.
cruises, 2020
versus 2026
The recent
cluster we see in the news is severe but they are limited cases from a north
Atlantic bound vessel, originating from South America. 5% case rate on the ship, resulting so
far in over 150 infections and a handful of deaths.
It was likely from rodent exposure onshore in Argentina region, and then
transmitted abroad. Government officials
have been able to contain the spread: remove affected passengers or quarantine
in place, and carry on with effective containment via tracing. Not dissimilar to current travel bug
protocols.
Contrast
that with the Diamond Princess in 2020.
Where a fifth of the similar-sized cruise was infected, or 712
infections. Most were quickly and asymptomatically
spread and resulted in nearly a dozen deaths.
Over time globally the fatality ratio from coronavirus winnowed to 1-2%
adjusted.
If the 2026
cruise had 2020 statistics, we’d have already reached nearly a thousand cases,
and multiple more already in a second wave with a multiple higher deaths. Yet this hadn’t occurred and the system has
contained itself, while sailing between lands, out in the ocean. However what is similar to 2020 is the
current small numbers make statistical analysis even for the cruise population erratic,
including for the death rate which is still emerging as we publish this
article.
Within this
comparison of just two ships, is also a lesson about selection bias skewing. We are only now seeing the published
official, and lagged, statistics for example on the first couple deaths from
hantavirus. And this leads to an overestimation
of the death rate because mild or asymptomatic cases that are still not yet
known nor tested. So too when a virus is
known only as it is "too deadly" too quickly [e.g., hantavirus or
Ebola], it often fails to reach pandemic proportions because it kills the host
before they can interact with a wide enough circle. So the proper sizing of
hantavirus needs to be scaled for the long term potential, adjusted for the
selection bias in the fatalities and spread, both over time.
back to Covid-19
There are
deep probability and logic lessons from our coronavirus analyses responded to even by the China government, and they break through lessons of
group think in linear models for epidemiology.
Our statistical emphasized survivorship bias, the reason for still
underappreciated multi-wave dynamics, and selection effects [e.g., who gets
exposed, who survives initial versus later waves, behavioral changes that
conditionalize from caution]. We should
note that no one in my family contracted coronavirus, though many in the
circles around me had.
We’ll need
to run parallel analysis here to better understand the very small number of early
severe cases [adjusted for the traveling population] that might reflect
higher-risk exposures or vulnerable individuals. And we have yet to see enough time series
data around the world for what would be defined as “later/milder” waves of
cases exposing survivors.
For those
interested in the even earlier pandemic proxy, the 1918 flu had waves with
varying lethality also partly due to viral evolution, population immunity, and
behavior. While there is no biological reason for multiple waves, there is a
statistical one. However hantavirus
doesn't evolve/transmit the same way, and so the waves of fading virus cases
throughout the survived and risk-taking population is fitting the endemic and low-reproduction
nature better than a novel respiratory nature.
broad
statistics philosophies
We are
looking at a sample [a cruise] of a small-sample of total global cases over a few
years. There is heavy selection bias at
play, and regional supportive ICU care for hantavirus patients, and low overall
transmissibility. While the data is too
early and immature globally at this stage, this will not be coronavirus level. It will frighten and lead to precautions
still as we began a busy summer season, and that’s normally included in
modeling predictions especially with coronavirus not being that long ago in
people’s memories.
But
probability modeling on this website and our social media elevates to looking rigorously
at the framing of uncertainty, and the asymmetric nature of predictive errors
particularly on upswings in cases.
Even as we
don’t see the hallmarks between the initial cruise cases in 2020 versus 2026,
the larger survivorship and wave dynamics will still apply, as will the copulas
of joint tails in severity, reach, and biased inherent responses. This may not even be a silver swan
event.
Also worth
noting some of the same public health leaders from 2020 era are still in
leadership roles today, and still with their flawed interpretations of virological
policy which in hindsight was not as well addressed in Western nations as in
the origin country itself.
We also see
the need to be careful how to treat uncertainty in group-thinking and
centralized leaders. This is the
difference between mathematical logic, and biology. Having a rare and independent, statistically
literate voice, continues to add much value to be preventative and not fall for
the protocols stemming from superficially scary similarities to a previous
pandemic.
Our core mathematical
insights are not just “it’s not Covid-19”, which is as this year’s version of
the 2020 “it’s just the flu” commentary.
There are deeper analogies:
[1] modern
systems are often vulnerable to fatter-tail events
[2] authorities
systematically underestimate nonlinear risk in their toolkit
[3] convex
or exponential spread is psychologically [and mathematically] invisible early
on
[4] early
data in outbreaks are distorted by severe ascertainment bias, lag, and
non-harmony across regions
[5] early
biased cases distort fatality ratios with lagging cohorts and
unequal\uncategorized medical access
[6] institutional
messaging tends to be corporate, lagged, and biasedly incentivized because
systems optimize for population stability, not individuals
This was one
of my most popular articles along with the corresponding public
prediction calculator: read by millions and used by thousands. The key thing
we grasped early in 2020, which many officials did not, was the optimal
combination of lethality and presymptomatic spread is what leads to exponential
and overwhelming civilization-reach. And
this is again a statistical insight, not a virology insight.
Medical
communities in January–February 2020 focused on: low apparent CFR, small
visible case counts, and comparisons to flu.
They misused math and lacked the ability to integrate into their
analysis the humility to consider: compounding, delayed feedback loops, network
dynamics, healthcare access and saturation effects, and lack of longitudinal
and consistent vaccine endpoint studies.
That was the
real signal lost among policy makers, for years later grounded in their
sunken-cost political fallacy.
|
disease |
fatality |
transmissibility |
outcome |
|
Ebola |
very high |
limited |
outbreaks |
|
SARS-1 |
high |
moderate |
contained |
|
Coronavirus |
moderate |
extremely high |
pandemic |
|
Hantavirus |
very high |
very low |
sporadic clusters |
“three-cycles” and survivorship
While not a
formal epidemic concept it is something we’ve noted from an advanced
probability modeling dynamic.
Consider the
selection bias first, as most vulnerable people may die quickly, early on, as
they spread prior to even showing symptoms.
The nursing homes scandalously left behind a wake of forgotten corpses
as we recall in New York 2020.
However
there are medicines, precautions, and also behavioral adaptation to a large
remaining population. Transit has become
halted, people worked from home more, and immunity accumulated as well within
the population. Hospitals improved
treatment but also had less severe later cases.
Urgent care centers and others were able to test as well later.
The virus continued to new combinations for the remaining population. It became more transmissible and lower
lethality. It also faced a more
resistant population that survived by that time for anything in such a wave to
even become noticed. Bear in mind too
the case counts that graced the cover of newspapers then showed spikes in cases
more so than spikes in fatalities, and so it would have been harder to isolate
mini-waves of lower transmission and high fatality strains within the evolution
timeline. And when the breakthrough
deaths started heading in an inconvenient convergence, the data series stopped.
The media is
a mirror of the survivorship. With the
early cases in all the previous outbreaks [1918, early coronavirus, SARS, many
hemorrhaging outbreaks] we are reporting what people fear which is the
deadliest cases. But in fact we may only
be seeing the ascertainment distortion mirrored back to us.
global
transmission?
It is
appealing to consider the mirage of a virus operating the same in every global region
where infections take hold. And yet over
time we see this has not occurred. And
perhaps it is too late for civilization-level transmission. The South America statistics may also be
confused with the same ascertainment and survivorship bias we see in the earlier
evolutions of other diseases, meaning the same initial demographics in other globes
may now fare different and in this case better.
The Andes-strain
in South America broke through an initial immunity herd there and it has not
yet been anywhere near as replicated, anywhere else on earth. It’s had a chance to however and the cruise
would have been such a petri example.
conclusion
To be sure,
the fatalities outside the Americas can still be higher than desired [even
though lower than the up to 50% in the Americas], but this could still lead to
broken down statistics and a lack of pandemic potential. If the replacement ratio doesn’t increase by
a significant amount going forward; something that appears near impossible and
unnatural, to now suddenly do.
Even in the
current travel clusters, such as airplanes, trains, stadiums, and other tight
travel and entertainment modes. Cluster
and duration in confined spaces is key. We
also don’t have a surge of global troop movement on the scale known during
1918. So such a modern transmission that
might facilitate global waves, also
doesn’t exist.
And it may have also been low when French actor Gene Hackman didn’t become infected or die of same a year ago, in a tight secluded isolation with his virus-infected wife. Whether macro or micro, over the long development in virus duration, we expect the spread to remains low globally from this point forward. Even as we see the 2026 cruise news, it may have less emotional impact in the future even with some continued cases over time: versus the emotional whiplash it has today.
KILLED THE STRONG BOND
No comments:
Post a Comment