Read on Twitter
*Thread* Why is the death toll so low? Is it true that we have the "lowest fatality rate in the world?"
No. Far from it. And really and truly, I know it *seems* easy to interpret data, but it& #39;s critical to listen to the experts... because it& #39;s *not* so simple. 4 reasons: (1/n)
No. Far from it. And really and truly, I know it *seems* easy to interpret data, but it& #39;s critical to listen to the experts... because it& #39;s *not* so simple. 4 reasons: (1/n)
1) The death toll is a lagging indicator. If you compare deaths to diagnosed cases just after a massive spike in cases, particularly for a disease that takes a few weeks to produce outcomes (outcomes = death, disability, recovery), you& #39;re doing it wrong. (2/n)
We have enough cases now to spot trends. We know that Black people, POC, certain age groups, diabetes, & hypertension make a person more likely to die. It& #39;s not hard to look at these new cases and see that many will die. It& #39;s just a matter of time. (3/n)
To truly assess our fatality rates, we have to look at cases that actually have outcomes. When we do that, we can see that we& #39;re not only losing a LOT of Americans; we& #39;re losing younger lives than many other countries. (4/n)
2) Lead Time Bias: The length of this lag *increases* as the general public becomes more aware. Contact tracing means earlier testing. Many are now being tested after exposure rather than several days after the onset of symptoms. (5/n)
This can look like we& #39;ve improved. Look at how many weeks have passed & the death toll is still low! No, the duration *increases* because of earlier diagnoses. That doesn& #39;t mean better outcomes. We& #39;ve prolonged the inevitable and celebrating now is short-sighted. (6/n)
3) Deaths tolls are increasing: One common mistake people make when checking the numbers is to hear the headlines about how Florida, Texas, and Arizona are skyrocketing, then look at the nation& #39;s death toll and assume deaths are equally distributed. (7/n)
The US is a big country. We& #39;re massive, and we have 50 independent states operating on their own terms and within those states we have numerous regions operating within their own terms. (8/n)
(This is also the problem with using personal experiences / anecdotes to draw conclusions at the larger level but I digress.) (9/n)
When we look at the hotspots, we can see that the death toll started increasing last week. FL, AZ, and TX are NOT ok. Correctly interpreting these data reveals that these states are heading straight into an unmitigated disaster.
(10/n)
(10/n)
4) Which brings us to Simpson& #39;s Paradox: This is worthy of its own thread but worth introducing. This occurs when you have clear trends between groups when contrasting certain strata, but then when you combine them, the differences disappear or even appear reversed. (11/n)
In other words, you can have two countries: Nation A and B. Nation A can have higher fatality rates among people in every single age group. But when you compare death rates among the Nation A and Nation B, Nation A appears to have lower overall death tolls. (12/n)
It seems impossible, but is actually pretty common. (I& #39;m telling you, interpreting statistics isn& #39;t a DIY project.) The way it occurs is related to the population distribution among ages in the above example, though in real life it& #39;s typically more complex than that. (13/n)
On every metric, we& #39;re failing. We& #39;re failing to contain the virus. We& #39;re failing to keep people alive after they become sick. COVID is one massive American failure and incorrectly citing the death toll as a sign of success is merely distracting us from what& #39;s ahead. (14/n)
tl:dr: We do not at all have the best fatality rates in the world; we& #39;re... as many public health professionals said earlier this week... literally in a free fall. (15/15)
